Apposition and Coordination in Australian Languages ... - CiteSeerX
Apposition and Coordination in Australian Languages ... - CiteSeerX
Apposition and Coordination in Australian Languages ... - CiteSeerX
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Apposition</strong> <strong>and</strong> <strong>Coord<strong>in</strong>ation</strong> <strong>in</strong> <strong>Australian</strong> <strong>Languages</strong>: An LFG Analysis<br />
Louisa Sadler <strong>and</strong> Rachel Nordl<strong>in</strong>ger<br />
University of Essex <strong>and</strong> University of Melbourne<br />
Abstract<br />
It is well known that <strong>Australian</strong> languages make heavy use of nom<strong>in</strong>al apposition.<br />
However there is little discussion <strong>in</strong> the theoretical literature of how such appositional<br />
structures should be analysed syntactically. We present a range of data from <strong>Australian</strong><br />
languages illustrat<strong>in</strong>g how multiple nom<strong>in</strong>als share a s<strong>in</strong>gle gramamtical function with<strong>in</strong><br />
the clause. We argue that such constructions should be treated as set-valued grammatical<br />
functions <strong>in</strong> LFG. Sets as values for functions are well-established <strong>in</strong> LFG <strong>and</strong> are<br />
used <strong>in</strong> the representation of adjuncts, <strong>and</strong> also <strong>in</strong> the representation of coord<strong>in</strong>ation. In<br />
many <strong>Australian</strong> languages, coord<strong>in</strong>ation is expressed asyndetically, that is, by nom<strong>in</strong>al<br />
juxtaposition with no overt coord<strong>in</strong>ator at all. We argue that the syntactic similarity of<br />
all of these juxtaposed constructions (coord<strong>in</strong>ations <strong>and</strong> appositions alike) motivates an<br />
analysis <strong>in</strong> which they are treated similarly <strong>in</strong> the syntax, but suitably dist<strong>in</strong>guished <strong>in</strong> the<br />
semantics. We show how this can be achieved with<strong>in</strong> LFG, provid<strong>in</strong>g a unified treatment<br />
of juxtaposition <strong>in</strong> <strong>Australian</strong> languages, <strong>and</strong> demonstrat<strong>in</strong>g the strength of the modular<br />
Lexical-Functional Grammar approach.<br />
1 Introduction<br />
<strong>Australian</strong> languages are characterised by a prevalence of appositional (juxtaposed) NP structures<br />
express<strong>in</strong>g a wide range of construction types, <strong>in</strong>clud<strong>in</strong>g pronoun-noun appositions, partwhole<br />
constructions, generic-specific constructions <strong>and</strong> <strong>in</strong>clusory constructions (e.g. Blake<br />
(1987)). 1 Such appositional nom<strong>in</strong>al structures <strong>in</strong> <strong>Australian</strong> languages have received very little<br />
attention <strong>in</strong> the theoretical literature. In fact, there is relatively little discussion of apposition<br />
<strong>in</strong> the theoretical literature at all (notable exceptions be<strong>in</strong>g de Vries (2006); Van Eynde (2005)),<br />
despite the fact that apposition <strong>in</strong> European languages has received a reasonable amount of descriptive<br />
attention (see for example, Quirk et al. (1985); Meyer (1992); Acuña-Fariña (1999)).<br />
In this paper we seek to redress this imbalance by provid<strong>in</strong>g an analysis of NP appositions <strong>in</strong><br />
<strong>Australian</strong> languages <strong>in</strong> Lexical-Functional Grammar (LFG), which exploits the flexible architecture<br />
of LFG to capture the syntactic similarities of the various juxtaposed NP constructions<br />
while allow<strong>in</strong>g them to be suitably dist<strong>in</strong>guished <strong>in</strong> the mapp<strong>in</strong>g to the semantics.<br />
<strong>Australian</strong> languages also commonly use juxtaposition to express NP coord<strong>in</strong>ation, mean<strong>in</strong>g<br />
that there is often little difference <strong>in</strong> the surface syntax between apposed <strong>and</strong> coord<strong>in</strong>ated<br />
NPs. Thus, we argue that these construction types – both apposition <strong>and</strong> coord<strong>in</strong>ation alike<br />
1 This paper results from jo<strong>in</strong>t research <strong>in</strong>to coord<strong>in</strong>ation strategies <strong>in</strong> <strong>Australian</strong> Aborig<strong>in</strong>al languages funded<br />
by a British Academy grant (SG-39545). Earlier versions of this work have been presented at the ALS05 Conference<br />
<strong>in</strong> Melbourne 2005, <strong>and</strong> the LFG06 conference <strong>in</strong> Konstanz, 2006. We thank both audiences for helpful<br />
feedback that has led to substantial improvements <strong>in</strong> the presentation <strong>and</strong> argumentation. Of course, we rema<strong>in</strong><br />
responsible for rema<strong>in</strong><strong>in</strong>g errors <strong>and</strong> <strong>in</strong>adequacies.<br />
1
– should be treated similarly <strong>in</strong> the syntax but differentiated <strong>in</strong> the mapp<strong>in</strong>g to the semantics;<br />
an approach that is possible us<strong>in</strong>g the modular framework of LFG. Our analysis builds<br />
on the st<strong>and</strong>ard LFG approach to coord<strong>in</strong>ation (Kaplan <strong>and</strong> Maxwell, 1988; Dalrymple <strong>and</strong><br />
Kaplan, 2000; Dalrymple, 2001) <strong>in</strong> which coord<strong>in</strong>ated NPs are treated as set-valued grammatical<br />
functions at f-structure. We argue that NP appositions, <strong>in</strong> which multiple nom<strong>in</strong>als<br />
<strong>in</strong> juxtaposition share a s<strong>in</strong>gle grammatical function <strong>in</strong> the clause, should likewise be treated<br />
as sets at f-structure, provid<strong>in</strong>g a unified analysis of a range of NP juxtapositions common to<br />
many <strong>Australian</strong> languages. The differences between the various appositional <strong>and</strong> coord<strong>in</strong>ative<br />
constructions are then captured <strong>in</strong> the mapp<strong>in</strong>g to the semantics, which we discuss <strong>in</strong> some<br />
detail <strong>in</strong> §6. Our approach not only provides a natural <strong>and</strong> unified account of NP juxtaposition<br />
<strong>in</strong> <strong>Australian</strong> languages, but suggests an approach to apposition more generally <strong>in</strong> which apposed<br />
constituents are treated as syntactically analogous to coord<strong>in</strong>ated constituents, captur<strong>in</strong>g<br />
the syntactic similarities between the construction types that have been noted <strong>in</strong> the literature<br />
on apposition <strong>in</strong> other languages such as English (Quirk et al., 1985; Meyer, 1992; de Vries,<br />
2006).<br />
2 <strong>Apposition</strong>al constructions<br />
The follow<strong>in</strong>g exemplify a range of different <strong>in</strong>terpretations associated with juxtaposed nom<strong>in</strong>al<br />
structures <strong>in</strong> <strong>Australian</strong> languages, all of which are frequently analysed as appositional<br />
constructions <strong>in</strong> <strong>Australian</strong> language descriptions (e.g. Blake (1979, 1983, 1987, 2001); Evans<br />
(1995); Heath (1978, 1984)). These constructions all have <strong>in</strong> common the fact that they <strong>in</strong>volve<br />
the juxtaposition of nom<strong>in</strong>al elements hav<strong>in</strong>g both the same referent <strong>and</strong> the same grammatical<br />
function, as evidenced by the fact that the nom<strong>in</strong>als <strong>in</strong>volved are all <strong>in</strong>flected for the same case<br />
feature. Furthermore, <strong>in</strong> all of these cases there is little evidence that one or other nom<strong>in</strong>al<br />
can be identified as the syntactic head of the construction. Obviously languages will differ <strong>in</strong><br />
terms of the range of constructions that they encode with NP juxtaposition, but those exemplified<br />
here are fairly typical. In (1) we see a generic-specific construction, with the generic noun<br />
‘elasmobranch’ juxtaposed to the specific noun ‘shark’; (2) exemplifies a part-whole construction<br />
with the whole nom<strong>in</strong>al (‘bundle’) juxtaposed to the part nom<strong>in</strong>al (‘fight<strong>in</strong>g stick’); 2 <strong>and</strong><br />
(3) <strong>and</strong> (4) illustrate two variants of straightforward appositional constructions – a nom<strong>in</strong>alnom<strong>in</strong>al<br />
appositional construction <strong>in</strong> (3) <strong>in</strong> which ‘old man’ is apposed to ‘husb<strong>and</strong>’ <strong>in</strong> subject<br />
function, 3 <strong>and</strong> a nom<strong>in</strong>al-pronom<strong>in</strong>al appositional construction <strong>in</strong> (4) <strong>in</strong> which the coord<strong>in</strong>ated<br />
NP ‘those men <strong>and</strong> women’ is apposed to the coreferential third person plural pronoun bi-l-da. 4<br />
2 These appositional nom<strong>in</strong>al structures are dist<strong>in</strong>guished from part-whole possessive constructions, which are<br />
also often expressed with juxtaposition of nom<strong>in</strong>als, but which clearly constitute a head-modifier construction.<br />
3 Note that the two apposed nom<strong>in</strong>als come before the auxiliary g<strong>in</strong>-amany here, show<strong>in</strong>g them to jo<strong>in</strong>tly<br />
belong to an NP constituent s<strong>in</strong>ce the Wambaya auxiliary must always be the second constituent <strong>in</strong> the clause<br />
(Nordl<strong>in</strong>ger 1998a).<br />
4 The abbreviations used <strong>in</strong> examples are: A ‘transitive subject’; ABS ‘absolutive’; ACC ‘accusative’; ACT<br />
‘actual’; APASS ‘antipassive’; ASS ‘associative’; CON ‘cont<strong>in</strong>ous’; DAT ‘dative’; DU ‘dual’; ERG ‘ergative’; EXC<br />
‘exclusive’; EX ‘exclusive’; FUT ‘future’; IMPF ‘imperfective aspect’; INC ‘<strong>in</strong>clusive’; INCH ‘<strong>in</strong>choative’; LOC<br />
‘locative’; MLOC ‘modal locative’; M ‘mascul<strong>in</strong>e’; NEUT ‘neuter’; NM ‘nom<strong>in</strong>aliser’; NOM ‘nom<strong>in</strong>ative’; NON-<br />
2
(1) Dath<strong>in</strong>-a dangka-a niya<br />
that-NOM man-NOM<br />
kulkiji-y.<br />
shark-MLOC<br />
3SG.NOM<br />
wumburung-kuru raa-ja<br />
spear-PROP<br />
wanku-ya<br />
spear-ACT elasmobranch-MLOC<br />
‘That man speared a shark with a spear.’ (Evans, 1995, 244: Kayardild)<br />
(2) kawuka<br />
bundle<br />
jardiyali<br />
fight<strong>in</strong>g.stick<br />
‘a bundle of fight<strong>in</strong>g sticks’ (ibid, 249: Kayardild)<br />
(3) Garidi-ni<br />
bungmanyi-ni<br />
husb<strong>and</strong>.I-ERG old.man.I-ERG<br />
g<strong>in</strong>-amany<br />
3SG.M.A-P.TWD<br />
yanybi.<br />
get<br />
‘(Her) old man husb<strong>and</strong> came <strong>and</strong> got (her).’ (Nordl<strong>in</strong>ger, 1998a, 133: Wambaya)<br />
(4) Dath<strong>in</strong>-a<br />
maku-wa<br />
bithi<strong>in</strong>-da bi-l-da<br />
that-NOM woman-NOM man-NOM<br />
warra-j.<br />
3-PL-NOM go-ACT<br />
‘Those men <strong>and</strong> women are go<strong>in</strong>g.’ (Evans, 1995, 249: Kayardild)<br />
<strong>Apposition</strong>al nom<strong>in</strong>al structures are extremely common <strong>in</strong> <strong>Australian</strong> languages, as noted by<br />
Blake (1987: 77) who states that juxtaposed nom<strong>in</strong>als referr<strong>in</strong>g to the same entity “may be <strong>in</strong><br />
parallel, each contract<strong>in</strong>g its own relationship with the verb.” Indeed, some researchers have<br />
gone so far as to claim for particular <strong>Australian</strong> languages that they have no noun phrases at<br />
all, with all nom<strong>in</strong>als exist<strong>in</strong>g <strong>in</strong> apposition. This is the analysis given by Blake (1983) for<br />
Kalkatungu, for example; <strong>and</strong> by Heath (1978) for Ng<strong>and</strong>i <strong>in</strong> which “noun phrases which have<br />
more than one constituent are typically formed by apposition” (p. 52). Even languages with<br />
clear NP structures for head-modifier relations have a range of appositional nom<strong>in</strong>al constructions.<br />
For example, Evans (1995: 247) <strong>in</strong> discuss<strong>in</strong>g the Kayardild constructions exemplified<br />
<strong>in</strong> (1) <strong>and</strong> (2) above, states that “there are no syntactic reasons for consider<strong>in</strong>g one nom<strong>in</strong>al to<br />
be the head, <strong>and</strong> it is better to treat them as apposed nom<strong>in</strong>als”.<br />
A further type of juxtaposed construction common to <strong>Australian</strong> languages is the <strong>in</strong>clusory<br />
construction (S<strong>in</strong>ger, 2001, 2005) (also known <strong>in</strong> the literature as the ‘plural pronoun<br />
construction’ (Schwartz, 1988a)), <strong>in</strong> which a plural pronoun referr<strong>in</strong>g to the superset is comb<strong>in</strong>ed<br />
with a subset nom<strong>in</strong>al. In many languages the <strong>in</strong>clusory construction <strong>in</strong>volves simple<br />
juxtaposition of the two elements, as <strong>in</strong> the follow<strong>in</strong>g from Kayardild:<br />
(5) Nga-rr-a<br />
1-DU-NOM<br />
kajakaja<br />
daddy.NOM<br />
warra-ja thaa-th.<br />
go-ACT<br />
return-ACT<br />
‘Daddy <strong>and</strong> I will go’ (lit. ‘We two, <strong>in</strong>clud<strong>in</strong>g daddy, will go’) (Evans, 1995, 249:<br />
Kayardild)<br />
SING ‘non-s<strong>in</strong>gular’; NP ‘non-past’; P ‘past’; PL ‘plural’; PRES ‘present’; PROP ‘proprietive’; RECIP ‘reciprocal’;<br />
SG ‘s<strong>in</strong>gular’; SUB ‘subject’; TWD ‘directions towards’; VE ‘vegetable’.<br />
3
Although <strong>in</strong>clusory constructions have much <strong>in</strong> common with the appositional constructions<br />
exemplified <strong>in</strong> (1) to (4) above, they differ <strong>in</strong> that the features of the whole construction are<br />
those of only one of the constituent parts – namely the superset pronoun. We return to a<br />
discussion of <strong>in</strong>clusory constructions <strong>and</strong> their relationship to appositional juxtapositions <strong>in</strong><br />
§7.<br />
Return<strong>in</strong>g to the appositional structures which are the focus of this section, we can characterise<br />
them as nom<strong>in</strong>al constructions <strong>in</strong> which one or more nom<strong>in</strong>al elements hav<strong>in</strong>g the same<br />
(or similar) referent serve together as a s<strong>in</strong>gle argument, with no clear evidence of syntactic<br />
dependency or subord<strong>in</strong>ation. <strong>Apposition</strong>s are therefore syntactically similar to coord<strong>in</strong>ations,<br />
differ<strong>in</strong>g only <strong>in</strong> the semantic requirement that appositional elements have similar reference,<br />
as reflected <strong>in</strong> the follow<strong>in</strong>g def<strong>in</strong>itions (Crystal 1997):<br />
apposition: A traditional term reta<strong>in</strong>ed <strong>in</strong> some models of GRAMMATICAL description<br />
for a sequence of units which are CONSTITUENTS at the same grammatical<br />
LEVEL, <strong>and</strong> which have an identity or similarity of REFERENCE (p. 24)<br />
coord<strong>in</strong>ation: A term <strong>in</strong> GRAMMATICAL analysis to refer to the process or result<br />
of l<strong>in</strong>k<strong>in</strong>g LINGUISTIC UNITS which are usually of equivalent SYNTACTIC status.<br />
(p. 93)<br />
In fact, <strong>in</strong> many of these <strong>Australian</strong> languages NP coord<strong>in</strong>ation is encoded us<strong>in</strong>g exactly the<br />
same juxtaposed structure, as we shall now see.<br />
3 <strong>Coord<strong>in</strong>ation</strong> by Juxtaposition<br />
In very many <strong>Australian</strong> languages nom<strong>in</strong>al coord<strong>in</strong>ation is also expressed by juxtaposition of<br />
nom<strong>in</strong>als, as <strong>in</strong> (6) - (9) below. <strong>Coord<strong>in</strong>ation</strong> <strong>in</strong> these languages is therefore asyndetic, that is,<br />
does not <strong>in</strong>volve a coord<strong>in</strong>at<strong>in</strong>g affix, particle or word.<br />
(6) niya<br />
heNOM<br />
kurrka-tha barruntha-ya<br />
take-ACT<br />
wuran-ki<br />
nguku-y.<br />
yesterday-LOC food-MLOC water-MLOC<br />
‘Yesterday he took (with him) food <strong>and</strong> water.’ (Evans, 1995, 250: Kayardild)<br />
(7) Gaj-ba<br />
eat-FUT<br />
ngurru manganyma yangaji.<br />
1PL.INC(NP) tucker.III(ACC) meat.I(ACC)<br />
‘Let’s eat the bread <strong>and</strong> meat.’ (Nordl<strong>in</strong>ger, 1998a, 257: Wambaya)<br />
(8) Ngayirni babi-rni<br />
1SG.ERG<br />
ny<strong>in</strong>awarra.<br />
this.way<br />
older.brother-ERG<br />
ngiji-ng<strong>in</strong>yi-nu kujkarrana yam<strong>in</strong>ju-nu, nyu-rruku<br />
see-1DL.EXC-did two(M) shoot<strong>in</strong>g.star-did 2SG-went<br />
My brother <strong>and</strong> I saw two shoot<strong>in</strong>g stars when you’d gone. (Pensalf<strong>in</strong>i, 2003, 178:<br />
J<strong>in</strong>gulu)<br />
4
(9) Pala-nga ngatu jarri-nya-p<strong>in</strong>ti-ngi,<br />
that-LOC stationary INCH-NM-ASS-LOC<br />
kangkuru-jirri waraja yalapara.<br />
kangaroo-DU one goanna.<br />
mima-nik<strong>in</strong>yi-yi<br />
wait.for-IMPF-3PL.SUB<br />
puluku,<br />
3DU.DAT<br />
kujarra<br />
two<br />
‘And there, on the f<strong>in</strong>ish<strong>in</strong>g l<strong>in</strong>e, the two kangaroos <strong>and</strong> one goanna waited for those<br />
two.’ (Sharp, 2004, 315: Nyangumarta)<br />
The juxtaposed constructions <strong>in</strong> these examples are dist<strong>in</strong>guished from appositional nom<strong>in</strong>als<br />
as exemplified <strong>in</strong> §2 only by virtue of the fact that the two nom<strong>in</strong>als do not have the same referent.<br />
In some examples, this is reflected <strong>in</strong> the verbal agreement morphology which agrees with<br />
the resolved features of plural (9) or dual (8). Compare these with the Wambaya appositional<br />
construction, repeated <strong>in</strong> (10) below, <strong>in</strong> which the auxiliary shows s<strong>in</strong>gular number agreement<br />
(g<strong>in</strong>-amany). If the auxiliary had dual number agreement here (gurl-amany), the NP would be<br />
<strong>in</strong>terpreted as a coord<strong>in</strong>ation (‘(her) husb<strong>and</strong> <strong>and</strong> the old man’).<br />
(10) Garidi-ni<br />
bungmanyi-ni<br />
husb<strong>and</strong>.I-ERG old.man.I-ERG<br />
g<strong>in</strong>-amany<br />
3SG.M.A-P.TWD<br />
yanybi.<br />
get<br />
‘(Her) old man husb<strong>and</strong> came <strong>and</strong> got (her).’ (Nordl<strong>in</strong>ger, 1998a, 133: Wambaya)<br />
Apart from this semantic difference, such asyndetic coord<strong>in</strong>ations appear to be structurally<br />
identical to the appositional constructions discussed <strong>in</strong> §2, suggest<strong>in</strong>g that all of these juxtaposed<br />
constructions – both appositions <strong>and</strong> coord<strong>in</strong>ations – should be treated as hav<strong>in</strong>g the<br />
same syntactic structure. In fact, the observation that apposition <strong>and</strong> coord<strong>in</strong>ation are syntactically<br />
similar is found elsewhere <strong>in</strong> the literature, for example by Quirk et. al. (1985), de<br />
Vries (2006) <strong>and</strong> Meyer (1992), who observes that “while there are clear semantic differences<br />
between the two relations [i.e. apposition <strong>and</strong> coord<strong>in</strong>ation – LS & RN], syntactically the<br />
relations are quite similar” (p. 45). This suggests that what is appropriate for appositional<br />
structures is an analysis as a multiply-headed structure <strong>in</strong> the syntax, <strong>in</strong> which each element<br />
is <strong>in</strong>dependently fulfill<strong>in</strong>g the same grammatical function <strong>in</strong> parallel, similar to that usually<br />
assumed for coord<strong>in</strong>ation. In §5 we show how this can be captured <strong>in</strong> Lexical Functional<br />
Grammar (LFG) by extend<strong>in</strong>g the st<strong>and</strong>ard analysis of coord<strong>in</strong>ation to account for the appositional<br />
NP structures <strong>in</strong> <strong>Australian</strong> languages as well, provid<strong>in</strong>g both a natural account of these<br />
constructions <strong>and</strong> a unified account of juxtaposition <strong>in</strong> <strong>Australian</strong> languages. First, however,<br />
we review the st<strong>and</strong>ard treatment of coord<strong>in</strong>ation <strong>in</strong> LFG <strong>and</strong> show how it can be extended to<br />
account for the asyndetic coord<strong>in</strong>ation structure exemplified above.<br />
4 LFG Analysis of <strong>Coord<strong>in</strong>ation</strong><br />
LFG is a lexicalist constra<strong>in</strong>t-based syntactic framework which posits two co-present levels<br />
of syntactic representation: c-structure, which encodes phrase structural relations (of precedence<br />
<strong>and</strong> dom<strong>in</strong>ance) between elements, <strong>and</strong> f-structure, which represents a level of <strong>in</strong>ternal<br />
5
syntactic structure based on grammatical functions, modell<strong>in</strong>g predicate argument relations.<br />
C-structures are represented by phrase structure trees of a familiar sort <strong>and</strong> f-structures are<br />
represented by attribute value matrices. C-structures <strong>and</strong> f-structures (correspond<strong>in</strong>g to a given<br />
str<strong>in</strong>g) are <strong>in</strong>terrelated by means of a mapp<strong>in</strong>g function φ, which places elements of c-structure<br />
(<strong>in</strong> the doma<strong>in</strong>) <strong>in</strong> correspondence with f-structures. The mapp<strong>in</strong>g is expressed by means of<br />
equations associated with lexical items <strong>and</strong> phrase structure nodes. The relation between cstructure<br />
<strong>and</strong> f-structure is many-to-one <strong>and</strong> onto (f-structures can arise which are not related<br />
to specific c-structure nodes, <strong>and</strong> more than one c-structure node can correspond to a s<strong>in</strong>gle<br />
f-structure). Of particular <strong>in</strong>terest is the f-structure which is the m<strong>in</strong>imal structure satisfy<strong>in</strong>g<br />
a collection of constra<strong>in</strong>ts associated with the (c-structure) of a given utterance (the m<strong>in</strong>imal<br />
model): this is the f-structure associated with that utterance. F-structures are subject to wellformedness<br />
constra<strong>in</strong>ts, notably those of completeness <strong>and</strong> coherence, which ensure that all<br />
<strong>and</strong> only the arguments of a predicate are present (as values of subcategorisable grammatical<br />
functions) <strong>in</strong> the f-structure projected by a predicate. For comprehensive <strong>in</strong>troductions to the<br />
formalism see Bresnan (2001); Dalrymple (2001); Falk (2001).<br />
The framework provides a highly modular <strong>and</strong> flexible framework for syntactic description:<br />
<strong>in</strong> particular it has proved to be sufficiently flexible to provide an appropriate tool for<br />
modell<strong>in</strong>g the syntax of widely divergent languages rang<strong>in</strong>g from the highly configurational to<br />
the radically non-configurational. A significant body of work has developed LFG analyses of<br />
phenomena <strong>in</strong> <strong>Australian</strong> Aborig<strong>in</strong>al languages (Simpson <strong>and</strong> Bresnan, 1983; Simpson, 1991;<br />
Aust<strong>in</strong> <strong>and</strong> Bresnan, 1996; Nordl<strong>in</strong>ger, 1998b).<br />
A simple sentence such as (11) would have the f-structure (12). F-structures are simple<br />
attribute-value matrixes, where values can be atomic or complex. A particular feature of LFG<br />
is the use of set-valued attributes (that is, the value of the attribute is a set of f-structures), as<br />
<strong>in</strong> the case of the grammatical function ADJunct <strong>in</strong> (12), alongside atomic <strong>and</strong> complex valued<br />
features (TENSE <strong>and</strong> SUBJ respectively <strong>in</strong> (12)).<br />
(11) Kim wept yesterday.<br />
(12) ⎡<br />
⎢<br />
⎣<br />
SUBJ<br />
ADJ<br />
PRED ‘KIM’ <br />
PRED ‘YESTERDAY’ <br />
TENSE PAST<br />
PRED ‘WEEP< SUBJ >’<br />
⎤<br />
⎥<br />
⎦<br />
When a property is asserted to hold of a set, how it behaves will depend on the nature of the<br />
property: the majority of properties are distributive <strong>and</strong> such a feature is an attribute of every<br />
member of the set. The grammatical functions are distributive, thus Kim is the SUBJ of both<br />
the f-structure of shout <strong>and</strong> of the f-structure of cry <strong>in</strong> the (set of) f-structures for Kim shouted<br />
<strong>and</strong> cried. As we will see below, agreement features <strong>and</strong> the CONJ feature are non-distributive:<br />
when a feature is non-distributive it <strong>and</strong> its value are a property of the set as a whole.<br />
6
A st<strong>and</strong>ard analysis of NP coord<strong>in</strong>ation <strong>in</strong> LFG (Dalrymple <strong>and</strong> Kaplan, 2000; Dalrymple,<br />
2001) assumes a c-structure coord<strong>in</strong>ation schema along the l<strong>in</strong>es of (13) <strong>in</strong> which each coord<strong>in</strong><strong>and</strong><br />
is def<strong>in</strong>ed as belong<strong>in</strong>g to a set-valued f-structure by virtue of the annotation ↓∈ ↑ . 5<br />
(13)<br />
NP → NP CONJ NP<br />
↓∈ ↑ ↓∈ ↑<br />
The representation of a coord<strong>in</strong>ate structure <strong>in</strong>volves a hybrid object, that is, a set which additionally<br />
may itself have properties alongside the elements or members of the set. Consider the<br />
representation of (14), given <strong>in</strong> (15):<br />
(14) John <strong>and</strong> I met.<br />
(15) ⎡<br />
⎢<br />
⎣<br />
PRED ‘MET’<br />
SUBJ ji:<br />
⎡ ⎧<br />
⎡<br />
⎢ ⎢<br />
⎢ j: ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
i:<br />
⎢<br />
⎣<br />
⎢ ⎪⎩<br />
⎢<br />
⎣<br />
PRED ‘JOHN’<br />
INDEX<br />
PRED ‘PRO’<br />
INDEX<br />
CONJ AND<br />
INDEX<br />
PERS 1<br />
NUM PL<br />
NUM SG<br />
PERS 3<br />
NUM SG<br />
PERS 1<br />
<br />
⎤⎫<br />
⎥<br />
⎦<br />
⎪⎬<br />
⎤<br />
⎥<br />
⎦<br />
⎪⎭<br />
In (15), each conjunct contributes an f-structure to the set of f-structures which is the value<br />
of the SUBJ attribute (labelled j <strong>and</strong> i <strong>in</strong> (15)). Additionally, the SUBJ (labelled ji) has some features<br />
which express properties of the set as a whole - <strong>in</strong> this example, the feature CONJ with the<br />
value AND <strong>and</strong> the INDEX feature which expresses the PNG agreement features of the coord<strong>in</strong>ate<br />
structure as a whole (as noted above, agreement features are taken to be non-distributive).<br />
A common pattern of agreement with coord<strong>in</strong>ate NPs <strong>in</strong>volves syntactic resolution, whereby<br />
the agreement features of all the conjuncts are taken <strong>in</strong>to account <strong>in</strong> ‘calculat<strong>in</strong>g’ the agreement<br />
features of the coord<strong>in</strong>ate structure as a whole. Dalrymple <strong>and</strong> Kaplan (2000) develop<br />
an approach to syntactic feature resolution us<strong>in</strong>g set valued agreement features <strong>and</strong> the simple<br />
operation of set union. Thus for example, if PERS features are represented as shown <strong>in</strong> (16), set<br />
union gives the st<strong>and</strong>ard resolution pattern for this feature. Features such as CONJ <strong>and</strong> INDEX<br />
⎤<br />
⎤ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎦ ⎥<br />
⎦<br />
5 This is an <strong>in</strong>stance of a more general coord<strong>in</strong>ation scheme which comb<strong>in</strong>es like constituents, phrasal or<br />
lexical.<br />
7
are non-distributive (i.e. resolv<strong>in</strong>g) while most features, <strong>in</strong>clud<strong>in</strong>g grammatical functions such<br />
as SUBJ <strong>and</strong> OBJ, as already noted, are distributive. 6<br />
(16) {S,H} (1ST) ∪ {H} (2ND) = {S,H} (1ST)<br />
{S,H} (1ST) ∪ {} (3RD) = {S,H} (1ST)<br />
{H} (2ND) ∪ {} (3RD) = {H} (2ND)<br />
{} (3RD) ∪ {} (3RD) = {} (3RD)<br />
Similarly, a two gender (M, F) system with resolution to the mascul<strong>in</strong>e works as follows (with<br />
MASC correspond<strong>in</strong>g to the set {M} <strong>and</strong> FEM to the empty set):<br />
(17) {M} (MASC) ∪ {M} (MASC) = {M} (MASC)<br />
{M} (MASC) ∪ {} (FEM) = {M} (MASC)<br />
{} (FEM) ∪ {} (FEM) = {} (FEM)<br />
As shown <strong>in</strong> (18), the coord<strong>in</strong>ation schema for NP coord<strong>in</strong>ation <strong>in</strong> a language with syntactic<br />
feature resolution <strong>in</strong>volves simple f-descriptions which ensure that the PERS <strong>and</strong> GEND features<br />
of each NP conjunct be a subset of the PERS <strong>and</strong> GEND features of the set (<strong>in</strong> the follow<strong>in</strong>g,<br />
IND abbreviates INDEX). 7<br />
(18) NP −→ NP<br />
CONJ NP<br />
↓ ∈ ↑<br />
↓ ∈ ↑<br />
(↓ IND PERS) ⊆ (↑ IND PERS) (↓ IND PERS) ⊆ (↑ IND PERS)<br />
(↓ IND GEND) ⊆ (↑ IND GEND) (↓ IND GEND) ⊆ (↑ IND GEND)<br />
Resolution of the NUM feature, on the other h<strong>and</strong>, is not purely syntactic, as shown by the<br />
mimimally contrast<strong>in</strong>g examples <strong>in</strong> (19), to which we return briefly <strong>in</strong> §8.<br />
(19) The president <strong>and</strong> chief executive are attend<strong>in</strong>g the meet<strong>in</strong>g <strong>in</strong> Beirut.<br />
The president <strong>and</strong> chief executive is attend<strong>in</strong>g the meet<strong>in</strong>g <strong>in</strong> Beirut.<br />
In LFG it is possible to associate a collection of f-descriptions (or equations) with a name <strong>in</strong> a<br />
template def<strong>in</strong>ition. The rule schema can then call the template. This simple <strong>and</strong> convenient<br />
abbreviatory device has a number of useful properties. For example, because templates can<br />
call other templates, they can be organised to express l<strong>in</strong>guistic generalisations succ<strong>in</strong>ctly<br />
(see Dalrymple et al. (2004) for further discussion). The PERS <strong>and</strong> GEND equations for NP<br />
coord<strong>in</strong>ation can be expressed <strong>in</strong> a template:<br />
6 Distributive features are def<strong>in</strong>ed as follows:<br />
For any distributive property P <strong>and</strong> set s, P(s) iff ∀f ∈ s.P(f).<br />
For any nondistributive property P <strong>and</strong> set s, P(s) iff P holds of s itself. (Dalrymple <strong>and</strong> Kaplan, 2000)<br />
7 The orig<strong>in</strong>al formulation of syntactic resolution <strong>in</strong> Dalrymple <strong>and</strong> Kaplan (2000) does not refer to INDEX<br />
but simply to the <strong>in</strong>dividual PERS <strong>and</strong> GEND agreement features. S<strong>in</strong>ce then, a dist<strong>in</strong>ction between two sets of<br />
agreement features, INDEX <strong>and</strong> CONCORD has been postulated, see K<strong>in</strong>g <strong>and</strong> Dalrymple (2004), <strong>and</strong> we have<br />
updated the treatment of syntactic resolution here to be consistent with this later work.<br />
8
(20) NP-CNJT: (↓ IND PERS) ⊆ (↑ IND PERS)<br />
(↓ IND GEND) ⊆ (↑ IND GEND)<br />
We can therefore represent (18) more simply as <strong>in</strong> (21), <strong>in</strong> which each NP has been<br />
associated with the coord<strong>in</strong>ation template <strong>in</strong> (20):<br />
(21) NP −→ NP<br />
↓ ∈ ↑<br />
4.1 Asyndetic <strong>Coord<strong>in</strong>ation</strong><br />
@NP-CNJT<br />
CONJ NP<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
As we saw <strong>in</strong> §3, NP coord<strong>in</strong>ation <strong>in</strong> <strong>Australian</strong> languages is frequently encoded by juxtaposition,<br />
without any explicit marker of coord<strong>in</strong>ation. It seems reasonable to assume that such<br />
coord<strong>in</strong>ate structures receive precisely the same syntactic treatment as (14) above, so that they<br />
differ only <strong>in</strong> the presence (or absence) of a coord<strong>in</strong>ator. This suggests the c-structure schema<br />
<strong>in</strong> (22) for nom<strong>in</strong>al coord<strong>in</strong>ation. In (22) X is a categorial metavariable rang<strong>in</strong>g over N <strong>and</strong><br />
NP, 8 otherwise it differs from the st<strong>and</strong>ard schema <strong>in</strong> (13) above only <strong>in</strong> lack<strong>in</strong>g a CONJ feature.<br />
We defer full discussion of the agreement constra<strong>in</strong>ts (for the non-distributive INDEX<br />
features) to §6, but assume for the moment that the agreement template NP-CNJT is associated<br />
with each daughter node, as <strong>in</strong> (21).<br />
(22) X −→ X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
, X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
The f-structure correspond<strong>in</strong>g to the coord<strong>in</strong>ated subject <strong>in</strong> (9) (repeated here) is then that <strong>in</strong><br />
(24), with resolved INDEX features but no CONJ feature <strong>in</strong> the outer f-structure.<br />
(23) Pala-nga ngatu jarri-nya-p<strong>in</strong>ti-ngi,<br />
that-LOC stationary INCH-NM-ASS-LOC<br />
kangkuru-jirri waraja yalapara.<br />
kangaroo-DU one goanna.<br />
mima-nik<strong>in</strong>yi-yi<br />
wait.for-IMPF-3PL.SUB<br />
puluku,<br />
3DU.DAT<br />
kujarra<br />
two<br />
‘And there, on the f<strong>in</strong>ish<strong>in</strong>g l<strong>in</strong>e, the two kangaroos <strong>and</strong> one goanna waited for those<br />
two.’ (Sharp, 2004, 315: Nyangumarta)<br />
8 It is often impossible to tell whether coord<strong>in</strong>ation has taken place at the phrasal or lexical level <strong>in</strong> some of the<br />
data we have, <strong>and</strong> both are certa<strong>in</strong>ly possible. In some cases, the presence of a demonstrative will sometimes make<br />
clear the level at which elements are juxtaposed, but <strong>in</strong> general, ‘bare’ Ns can have <strong>in</strong> any case have referential<br />
NP mean<strong>in</strong>gs <strong>and</strong> constitute a full NP on their own.<br />
9
(24) ⎡<br />
⎢ ⎧ ⎡<br />
⎢ ⎢<br />
⎢ ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
⎣<br />
⎣<br />
⎪⎩<br />
INDEX<br />
<br />
PERS 3<br />
<br />
NUM PL<br />
PRED ‘GOANNA’<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
⎤<br />
⎥<br />
⎦<br />
PRED ‘KANGAROO’<br />
<br />
NUM DUAL<br />
INDEX<br />
PERS 3<br />
⎤<br />
⎫<br />
⎪⎬<br />
⎥<br />
⎦<br />
⎪⎭<br />
⎤<br />
⎥<br />
⎦<br />
Another <strong>in</strong>terest<strong>in</strong>g feature of NP coord<strong>in</strong>ation <strong>in</strong> <strong>Australian</strong> languages, consistent with the<br />
freedom of word order often characteristic of these languages, is that coord<strong>in</strong>ation structures<br />
can be discont<strong>in</strong>uous, as <strong>in</strong> the follow<strong>in</strong>g examples. That these are truly coord<strong>in</strong>ations (<strong>and</strong> not<br />
afterthoughts, for example) is readily established by such th<strong>in</strong>gs as discourse context (Blake,<br />
2001) <strong>and</strong> the fact that the verb shows resolved agreement (as <strong>in</strong> (25)).<br />
(25) K<strong>in</strong>tja-(ng)ku=yana<br />
female-ERG=<strong>and</strong><br />
(26) Ngul<br />
then<br />
<strong>in</strong>tji-mi-ngi-yu<br />
ntiya-(ng)ku tjipa-yi<br />
pelt-FUT-me-they.DU stone-ERG<br />
kurlay<strong>in</strong>gu-thu.<br />
this-ERG male-ERG<br />
‘The girl <strong>and</strong> boy will both pelt me with stones.’ (Blake, 2001, 423: Kalkatungu)<br />
ngay kirk kempthe kal-m thul=yuk<br />
1SG(ERG) spear(ACC) apart carry-P.IPFV woomera(ACC)<br />
‘I used to carry spears <strong>and</strong> woomeras separately’ (Gaby, 2006, 320: Kuuk Thaayorre)<br />
Our analysis will comb<strong>in</strong>e straightforwardly with the st<strong>and</strong>ard approach to discont<strong>in</strong>uity <strong>and</strong><br />
non-configurationality <strong>in</strong> LFG to account for these cases. In LFG c-structure nodes are generally<br />
optional – by the Pr<strong>in</strong>ciple of Economy, only those nodes which are motivated by overt lexical<br />
material (or by some semantic requirement) are present (Bresnan, 2001).<br />
In an example like (26) the two parts of the coord<strong>in</strong>ate structure appear separated by the<br />
verb <strong>and</strong> its modifier: each is represented at c-structure as a coord<strong>in</strong>ate structure with just one<br />
daughter present (27).<br />
(27) S<br />
NP<br />
(↑ SUBJ) = ↓<br />
N<br />
↑ = ↓<br />
I<br />
NP<br />
(↑ OBJ) = ↓a<br />
NP<br />
↓c ∈ ↑<br />
spear<br />
NP<br />
↓∈ (↑ ADJ)<br />
N<br />
↑ = ↓<br />
apart<br />
10<br />
V<br />
↑ = ↓<br />
carry<br />
NP<br />
(↑ OBJ) = ↓b<br />
NP<br />
↓d ∈ ↑<br />
woomera
In non-configurational structures, the grammatical functions to be assigned to an element<br />
of c-structure are not identified <strong>in</strong> configurational terms but by morphological means. In the<br />
example at h<strong>and</strong>, it is the case mark<strong>in</strong>g on the nom<strong>in</strong>als which <strong>in</strong>dicates the SUBJ (marked<br />
ERG) <strong>and</strong> the (discont<strong>in</strong>uous) OBJ (marked ACC). S<strong>in</strong>ce <strong>in</strong> pr<strong>in</strong>ciple, a wide range of different<br />
annotations might be compatible with a str<strong>in</strong>g of categories we assume NP under S are freely<br />
annotated (↑ GF) = ↓where GF is a meta-variable over the relevant set of grammatical functions.<br />
It is the case markers which identify which grammatical function a nom<strong>in</strong>al element maps to<br />
(Simpson, 1991; Aust<strong>in</strong> <strong>and</strong> Bresnan, 1996; Nordl<strong>in</strong>ger, 1998c).<br />
The two ACC-marked NPs under S will both <strong>in</strong>dependently co-specify the OBJ f-structure<br />
(as shown by the labels a <strong>and</strong> b <strong>in</strong> (27), <strong>and</strong> contribute elements to the (coord<strong>in</strong>ate) set.<br />
(28) ⎡<br />
PRED ‘CARRY’<br />
⎢ ADJ {[ PRED APART ]}<br />
⎢ ⎡ ⎧ ⎡<br />
⎢<br />
PRED<br />
⎢<br />
‘SPEAR’<br />
⎢ ⎢<br />
⎢ ⎢ ⎢<br />
⎢ ⎢ c: ⎣<br />
INDEX<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎪⎨<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎡<br />
⎢ OBJ a, b: ⎢<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎢<br />
d:<br />
⎢ ⎢<br />
⎣<br />
⎢ ⎢<br />
⎪⎩<br />
INDEX<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎢ ⎢ <br />
⎢ ⎣<br />
PERS 3<br />
INDEX<br />
⎢<br />
NUM PL<br />
⎢ ⎡<br />
⎤<br />
⎢<br />
PRED ‘PRO’<br />
⎢ ⎢ <br />
SUBJ<br />
⎣ ⎣<br />
NUM SG ⎥<br />
⎦<br />
INDEX<br />
PERS 1<br />
NUM SG<br />
PERS 3<br />
PRED ‘WOOMERA’<br />
NUM SG<br />
PERS 3<br />
⎤⎫<br />
⎥<br />
⎦<br />
⎪⎬<br />
⎤<br />
⎥<br />
⎦<br />
⎪⎭<br />
If we choose to associate some other annotation with the ACC marked NPs (treat<strong>in</strong>g one or<br />
both of them as contribut<strong>in</strong>g the whole OBJ function, for example), no complete <strong>and</strong> coherent<br />
f-structure would be produced. Our analysis of asyndetic coord<strong>in</strong>ation thus extends to account<br />
straightforwardly for the possibility of discont<strong>in</strong>uous coord<strong>in</strong>ations <strong>in</strong> these languages.<br />
5 Analysis of <strong>Apposition</strong><br />
Our proposal is that appositional nom<strong>in</strong>al structures should be treated as syntactically identical<br />
to coord<strong>in</strong>ations – namely, as hybrid structures at f-structure – with the differences between<br />
them captured <strong>in</strong> the mapp<strong>in</strong>g to the semantics. Representation as members of a set directly<br />
captures the <strong>in</strong>tuition about appositional structures: that they are co-heads, <strong>and</strong> do not st<strong>and</strong><br />
11<br />
⎤<br />
⎥<br />
⎤ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎥ ⎥<br />
⎦ ⎥<br />
⎦
<strong>in</strong> a syntactic dependency relation with one another. Furthermore, the fact that both coord<strong>in</strong>ations<br />
<strong>and</strong> appositions are treated similarly <strong>in</strong> the syntax captures the structural similarities<br />
between them that have been observed <strong>in</strong> the literature (e.g. Quirk et al (1985), Meyer(1992),<br />
de Vries(2006)). The various constructions may differ at f-structure, as we shall see, <strong>in</strong> terms<br />
of the agreement features of the set (i.e. whether they <strong>in</strong>volve feature resolution or not), <strong>and</strong><br />
then are further differentiated <strong>in</strong> the mapp<strong>in</strong>g to the semantic structure.<br />
On this view, the f-structure correspond<strong>in</strong>g to the apposition <strong>in</strong> (3), repeated <strong>in</strong> (3), is<br />
as <strong>in</strong> (30). Apart from the value of the non-distributive (INDEX) features of the set (which<br />
we discuss below), the f-structure <strong>in</strong> (30) is structurally identical to that associated with the<br />
coord<strong>in</strong>ation <strong>in</strong> (9) ((24) repeated as (31)).<br />
(29) Garidi-ni<br />
bungmanyi-ni<br />
husb<strong>and</strong>.I-ERG old.man.I-ERG<br />
g<strong>in</strong>-amany<br />
3SG.M.A-P.TWD<br />
yanybi.<br />
get<br />
‘(Her) old man husb<strong>and</strong> came <strong>and</strong> got (her).’ (Nordl<strong>in</strong>ger, 1998a, 133: Wambaya)<br />
(30) <strong>Apposition</strong>: ⎡<br />
⎢ ⎧ ⎡<br />
⎢ ⎢<br />
⎢ ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
⎣<br />
⎣<br />
⎪⎩<br />
(31) <strong>Coord<strong>in</strong>ation</strong>: ⎡<br />
INDEX<br />
<br />
PERS 3<br />
<br />
NUM SG<br />
PRED ‘HUSBAND’<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
PRED ‘OLD.MAN’<br />
INDEX<br />
⎢ ⎧ ⎡<br />
⎢ ⎢<br />
⎢ ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
⎣<br />
⎣<br />
⎪⎩<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
<br />
PERS 3<br />
<br />
NUM PL<br />
PRED ‘GOANNA’<br />
INDEX<br />
⎤⎫<br />
⎥<br />
⎦<br />
⎪⎬<br />
⎤<br />
NUM SG<br />
PERS 3<br />
⎥<br />
⎦<br />
⎪⎭<br />
⎤<br />
⎥<br />
⎦<br />
PRED ‘KANGAROO’<br />
<br />
NUM DUAL<br />
INDEX<br />
PERS 3<br />
⎤<br />
⎥<br />
⎦<br />
⎤<br />
⎫<br />
⎪⎬<br />
⎥<br />
⎦<br />
⎪⎭<br />
This analysis directly reflects the fact that there is no visible syntactic dist<strong>in</strong>ction with<strong>in</strong> the<br />
nom<strong>in</strong>al str<strong>in</strong>gs themselves between nom<strong>in</strong>al coord<strong>in</strong>ation <strong>and</strong> nom<strong>in</strong>al apposition. 9 In fact,<br />
9 Note that we are concerned here only with the lack of syntactic differences. There may well be <strong>in</strong>tonational<br />
or other differences between the two construction types.<br />
12<br />
⎤<br />
⎥<br />
⎦
as discussed <strong>in</strong> §2 above, the nom<strong>in</strong>al phrase <strong>in</strong> (29) is itself ambiguous between a coord<strong>in</strong>ative<br />
<strong>and</strong> an appositional <strong>in</strong>terpretation, disambiguated only by the verbal morphology. The<br />
auxiliary form g<strong>in</strong>-amany ‘3SG.M.A-P.TWD’ determ<strong>in</strong>es that the SUBJ is 3SG. If this example<br />
meant ‘the old man <strong>and</strong> her husb<strong>and</strong> (they)....’ then the f<strong>in</strong>ite auxiliary would be encoded with<br />
3DU. Crucially, the formal differences lie only <strong>in</strong> the agreement features of the set; there is<br />
no visible syntactic dist<strong>in</strong>ction with<strong>in</strong> the nom<strong>in</strong>al structure itself. Thus, as far as the syntax<br />
is concerned, our analysis needs to be able to account for the fact that the same nom<strong>in</strong>al<br />
f-structure may sometimes <strong>in</strong>volve feature resolution (i.e. <strong>in</strong> a coord<strong>in</strong>ation structure), <strong>and</strong><br />
sometimes not (i.e. <strong>in</strong> an appositional structure).<br />
Thus, we propose that both appositional <strong>and</strong> coord<strong>in</strong>ative constructions are licensed by the<br />
basic phrase structure schema <strong>in</strong> (32), with different annotations depend<strong>in</strong>g on issues of feature<br />
resolution <strong>and</strong> semantics, as discussed <strong>in</strong> detail <strong>in</strong> the follow<strong>in</strong>g section. <strong>Apposition</strong>al <strong>and</strong><br />
coord<strong>in</strong>ate structures differ syntactically, on this view, purely <strong>in</strong> terms of the overall agreement<br />
features of the structure as a whole. Structurally, they are identical.<br />
(32) X −→ X<br />
↓ ∈ ↑<br />
, X<br />
↓ ∈ ↑<br />
F<strong>in</strong>ally, note that all of the apposition types (generic-specfic, st<strong>and</strong>ard appositional, part-whole<br />
construction) can be also discont<strong>in</strong>uous, as exemplified with the generic-specific construction<br />
<strong>in</strong> (33).<br />
(33) Ngayika<br />
I<br />
ati-ntji<br />
meat-DAT<br />
ari-li<br />
eat-APASS<br />
thuwarr-ku.<br />
snake-DAT<br />
‘I’m eat<strong>in</strong>g snake.’ (Blake, 2001, 419: Kalkatungu)<br />
This will follow directly from the analysis of discont<strong>in</strong>uous coord<strong>in</strong>ate constructions discussed<br />
<strong>in</strong> §4 above. Namely we assume that each element <strong>in</strong> the c-structure rule is optional, under the<br />
Pr<strong>in</strong>ciple of Economy of Expression (Bresnan, 2001, 91), thereby allow<strong>in</strong>g each nom<strong>in</strong>al to<br />
constitute an NP on its own <strong>in</strong> the c-structure, while still contribut<strong>in</strong>g to a set at f-structure.<br />
6 Comput<strong>in</strong>g Mean<strong>in</strong>gs: Semantic Composition<br />
6.1 Coord<strong>in</strong>ate Mean<strong>in</strong>gs<br />
As we have seen, nom<strong>in</strong>al juxtapositions can have coord<strong>in</strong>ate mean<strong>in</strong>gs, <strong>in</strong>volv<strong>in</strong>g syntactic<br />
feature resolution <strong>and</strong> the construction of a coord<strong>in</strong>ate semantics. Coord<strong>in</strong>ate agreement can<br />
be captured <strong>in</strong> the template (20), repeated below as (34), follow<strong>in</strong>g Dalrymple <strong>and</strong> Kaplan<br />
(2000), which will suffice for present purposes as our focus is not on the details of agreement<br />
13
itself. 10 This template is associated with each constituent <strong>in</strong> the phrase structure rule, as <strong>in</strong><br />
(35), where we have coord<strong>in</strong>ate read<strong>in</strong>gs.<br />
(34) NP-CNJT: (↓ IND PERS) ⊆ (↑ IND PERS)<br />
(↓ IND GEND) ⊆ (↑ IND GEND)<br />
(35) X −→ X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
, X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
We now turn to the semantics, <strong>and</strong> <strong>in</strong> particular to semantic composition, that is, how mean<strong>in</strong>gs<br />
are associated with (<strong>and</strong> derived from) the syntactic structures we are discuss<strong>in</strong>g. It is<br />
generally assumed <strong>in</strong> LFG that semantic composition <strong>in</strong>volves reference to f-structures, that<br />
is, to representations of syntactic predicate-argument structure, rather than c-structures. Thus<br />
despite the widely vary<strong>in</strong>g c-structures that languages have, their semantics is largely <strong>in</strong>variant,<br />
<strong>in</strong> that the same sorts of mean<strong>in</strong>gs are encoded by a diverse range of external syntactic<br />
structures. 11<br />
The predom<strong>in</strong>ant approach to semantic composition <strong>in</strong> LFG, which we adopt here, uses<br />
l<strong>in</strong>ear logic for mean<strong>in</strong>g assembly. The mean<strong>in</strong>gs themselves will be represented as simple<br />
predicate logic expressions. These can be viewed as abbreviatory shorth<strong>and</strong> for more elaborate<br />
mean<strong>in</strong>g expressions. Our focus here is primarily on the process of mean<strong>in</strong>g assembly, that<br />
is, how mean<strong>in</strong>gs are associated with syntactic structures, <strong>and</strong> <strong>in</strong> particular on the syntaxsemantics<br />
<strong>in</strong>terface.<br />
In the glue (l<strong>in</strong>ear logic) theory of semantic composition <strong>in</strong>structions for comb<strong>in</strong><strong>in</strong>g mean<strong>in</strong>gs<br />
are stated as premises <strong>in</strong> a logical deduction. The order of composition is therefore determ<strong>in</strong>ed<br />
only as the logic itself determ<strong>in</strong>es it (<strong>and</strong> thus details of syntactic structure, such as the<br />
phrasal composition, are not important). A crucial aspect of the use of glue, or l<strong>in</strong>ear logic,<br />
for mean<strong>in</strong>g assembly is its resource sensitivity: once a premise is used, it is used up <strong>and</strong> no<br />
longer available for subsequent steps <strong>in</strong> the deduction. The approach is therefore motivated by<br />
the observed resource sensitivity of natural language <strong>in</strong>terpretation. A highly accessible <strong>in</strong>troduction<br />
to mean<strong>in</strong>g composition <strong>in</strong> LFG is provided by Dalrymple (2001, chapter 9), on which<br />
this brief outl<strong>in</strong>e is based. In this approach mean<strong>in</strong>g constructors are associated primarily with<br />
lexical items (though they can also be associated with phrase structure nodes <strong>in</strong> rules).<br />
Mean<strong>in</strong>g constructors have two parts: they are made up of a mean<strong>in</strong>g on the left h<strong>and</strong><br />
side <strong>and</strong> a logical formula over semantic structures correspond<strong>in</strong>g to that mean<strong>in</strong>g on the right<br />
10 Many <strong>Australian</strong> languages dist<strong>in</strong>guish four or more genders (e.g. MA, FEM, NEUT, VEG) <strong>and</strong> exhibit defaults<br />
<strong>and</strong> underspecification <strong>in</strong> gender agreement, <strong>and</strong> thus it is likely that several <strong>in</strong>terest<strong>in</strong>g issues will arise <strong>in</strong> the<br />
formulation of the details of gender resolution <strong>and</strong> gender agreement <strong>in</strong> such languages. For an approach to<br />
gender us<strong>in</strong>g complex GEND features rather than set-valued features, see Dalrymple et al. (2007).<br />
11 Of course this is not to deny that other sources of l<strong>in</strong>guistic <strong>in</strong>formation, or other types of l<strong>in</strong>guistic structure,<br />
are relevant to the process of semantic composition <strong>and</strong> <strong>in</strong>terpretation. The modular correspondence-based architecture<br />
of LFG allows non-syntactic levels of representation such as prosodic structure <strong>and</strong> <strong>in</strong>formation structure<br />
(which models discourse relations) to provide important <strong>in</strong>formation guid<strong>in</strong>g semantic <strong>in</strong>terpretation. Constra<strong>in</strong>ts<br />
aris<strong>in</strong>g through multiple projections may comb<strong>in</strong>e to determ<strong>in</strong>e the mean<strong>in</strong>g of a given utterance.<br />
14
h<strong>and</strong> side (semantic structures are related to f-structures by the projection σ). Consider as<br />
an example the mean<strong>in</strong>g constructor associated with the lexical entry for yawned given <strong>in</strong><br />
Dalrymple (2001, 233) <strong>and</strong> shown <strong>in</strong> (36).<br />
(36) yawned, V: (↑ PRED) = ‘YAWN < SUBJ > ’<br />
λX.yawn(X) : (↑ SUBJ)σ ⊸ ↑ σ<br />
The mean<strong>in</strong>g side of the constructor <strong>in</strong> (36) gives the mean<strong>in</strong>g of yawn as one-place predicate<br />
(of course, other more complex mean<strong>in</strong>gs can be substituted for FOPL expressions if required).<br />
Mean<strong>in</strong>g expressions are typed (so that the constant Kim, for example, is of type e): type<br />
<strong>in</strong>formation will determ<strong>in</strong>e how mean<strong>in</strong>g expressions comb<strong>in</strong>e with others. The glue side uses<br />
l<strong>in</strong>ear implication: (⊸) is an implication which can be read as say<strong>in</strong>g that if the mean<strong>in</strong>g<br />
resource for the SUBJ is available then it can be consumed to produce the mean<strong>in</strong>g of the<br />
sentence. 12 F<strong>in</strong>ally it is st<strong>and</strong>ard practice to <strong>in</strong>troduce labels as names of mean<strong>in</strong>g constructors<br />
to ease reference to them, a practice we will adopt. Thus the mean<strong>in</strong>g constructor for the<br />
<strong>in</strong>dividual Kim might be labelled as Kim :<br />
(37) Kim Kim: ↑ σ<br />
As for the semantics of NP coord<strong>in</strong>ation, Dalrymple (2001) associates the semantic contribution<br />
g-<strong>and</strong> (group-form<strong>in</strong>g <strong>and</strong>) <strong>in</strong> (38) with the coord<strong>in</strong>ator <strong>and</strong> <strong>in</strong> its lexical entry (40). In<br />
the mean<strong>in</strong>g constructor, recall that the left h<strong>and</strong> side is a mean<strong>in</strong>g expression <strong>and</strong> the right<br />
h<strong>and</strong> side is a glue constructor for resource sensitive semantic construction us<strong>in</strong>g l<strong>in</strong>ear implication.<br />
On the mean<strong>in</strong>g side, the lambda expression denotes the group form<strong>in</strong>g function - here<br />
a function from two <strong>in</strong>dividuals to the group conta<strong>in</strong><strong>in</strong>g those <strong>in</strong>dividuals. On the glue side,<br />
g-<strong>and</strong> consumes the semantics of one conjunct (↑ ∈)σ <strong>and</strong> produces a function from the semantics<br />
of the other conjunct to the semantics of the coord<strong>in</strong>ate structure as a whole. For more<br />
than b<strong>in</strong>ary coord<strong>in</strong>ation, a further semantic contribution g-<strong>and</strong>2, <strong>in</strong>volv<strong>in</strong>g the ! (of course)<br />
operator, can be used any number of times (<strong>in</strong>clud<strong>in</strong>g zero), each time add<strong>in</strong>g an <strong>in</strong>dividual<br />
<strong>in</strong>to the group (39). S<strong>in</strong>ce our primary focus here is on b<strong>in</strong>ary coord<strong>in</strong>ation, we will have no<br />
more to say about coord<strong>in</strong>ation <strong>in</strong>volv<strong>in</strong>g multiple conjuncts.<br />
(38) g-<strong>and</strong> λX.λY. { X, Y } :<br />
(↑ ∈)σ ⊸ [(↑ ∈)σ ⊸ ↑ σ ]<br />
(39) g-<strong>and</strong>2 λX.λY. { X } ∪ Y :<br />
!(↑ ∈)σ ⊸ [ ↑ σ ⊸ ↑ σ ]<br />
(40) <strong>and</strong> (↑ CONJ) = AND<br />
[g-<strong>and</strong>]<br />
[g-<strong>and</strong>2]<br />
12 As well as l<strong>in</strong>ear implication, l<strong>in</strong>ear logic uses so-called multiplicative conjunction (⊗), a form of conjunction,<br />
<strong>and</strong> the of course operator (!), which permits a premise to be used without be<strong>in</strong>g consumed.<br />
15
Follow<strong>in</strong>g Dalrymple (2001) <strong>and</strong> the arguments presented there<strong>in</strong>, we take it that NP coord<strong>in</strong>ation<br />
is correctly characterised <strong>in</strong> this way as group formation. However <strong>in</strong> the case of asyndetic<br />
coord<strong>in</strong>ation, there is no coord<strong>in</strong>ator <strong>in</strong> the structure with which to associate the semantics of<br />
g-<strong>and</strong>.<br />
Notice also that <strong>in</strong> languages (such as these) with three number dist<strong>in</strong>ctions (s<strong>in</strong>gular, dual<br />
<strong>and</strong> plural), it is not possible simply to associate the use of the group-form<strong>in</strong>g semantics with<br />
NUM resolution to PL, because the syntactic NUM of a group conta<strong>in</strong><strong>in</strong>g just a pair is DU. For<br />
present purposes, we restrict ourselves to b<strong>in</strong>ary coord<strong>in</strong>ation, <strong>and</strong> def<strong>in</strong>e the NUM resolution<br />
as <strong>in</strong> the template <strong>in</strong> (41). This captures the generalisation that either the overall number is<br />
DU (i.e. when two s<strong>in</strong>gular nom<strong>in</strong>als are coord<strong>in</strong>ated) or (at least) one of the constituents is<br />
non-s<strong>in</strong>gular, <strong>in</strong> which case the overall number is PL. Note that s<strong>in</strong>ce this makes reference to<br />
elements of the (coord<strong>in</strong>ate) f-structure, rather that to c-structure daughters, NUM resolution<br />
will operate correctly <strong>in</strong> cases of discont<strong>in</strong>uous coord<strong>in</strong>ation as well as cases of contiguous<br />
coord<strong>in</strong>ation by simple juxtaposition.<br />
(41) BINARY: {(↑ ∈ INDEX NUM) = SG ∧ (↑ INDEX NUM) = PL}<br />
| (↑ INDEX NUM) = DUAL<br />
To complete the <strong>in</strong>terpretation of nom<strong>in</strong>al juxtapositions as coord<strong>in</strong>ative, we need to associate<br />
the template BINARY <strong>and</strong> the mean<strong>in</strong>g constructor g-<strong>and</strong> with the phrase structure rule <strong>in</strong><br />
(35) (restrict<strong>in</strong>g attention to cases of b<strong>in</strong>ary coord<strong>in</strong>ation). S<strong>in</strong>ce there is no coord<strong>in</strong>ator, <strong>in</strong><br />
asyndetic coord<strong>in</strong>ation we arbitrarily associate them with one of the daughter constituents.<br />
The phrase structure rule for juxtaposed NPs <strong>in</strong> coord<strong>in</strong>ation, is therefore that <strong>in</strong> (42). 13<br />
(42) X −→ X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
@BINARY<br />
g-<strong>and</strong><br />
, X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
Our analysis of the juxtapositions with coord<strong>in</strong>ate semantics is thus analogous to the analysis<br />
of (non-juxtaposed) coord<strong>in</strong>ate constructions <strong>in</strong> other languages (Dalrymple <strong>and</strong> Kaplan 2001,<br />
Dalrymple 2001). In the next section we see how this same general approach can also provide<br />
an analysis of appositional juxtapositions.<br />
6.2 <strong>Apposition</strong>al Mean<strong>in</strong>gs<br />
In appositional juxtapositions the juxtaposed constituents are co-referential <strong>and</strong> there is no<br />
feature resolution at the level of the set: the features of the set are the same as the features<br />
13 In cases of discont<strong>in</strong>ous coord<strong>in</strong>ation, if neither part of the discont<strong>in</strong>uous structure is associated with these<br />
annotations, mean<strong>in</strong>g construction will fail, whereas if both parts are associated with these annotations, mean<strong>in</strong>g<br />
construction will also fail because there will be unconsumed premises. Therefore noth<strong>in</strong>g more needs to be added<br />
to ensure that only the right comb<strong>in</strong>ation of annotations is selected.<br />
16
of each of the members. Thus, <strong>in</strong> our terms, appositional constructions generally <strong>in</strong>volve<br />
INDEX shar<strong>in</strong>g between the set <strong>and</strong> the members of the set, as well as the construction of an<br />
appositional semantics.<br />
In order to capture the shar<strong>in</strong>g of INDEX features between the set members <strong>and</strong> the set<br />
itself, we def<strong>in</strong>e the appositional template <strong>in</strong> (43), which is associated with each of the daughter<br />
constituents <strong>in</strong> the appositional phrase structure rule, as <strong>in</strong> (44). This template ensures that the<br />
INDEX features of each daughter constituent are shared with the INDEX features of the set (i.e.<br />
a set conta<strong>in</strong><strong>in</strong>g two 3SG daughters will likewise have 3SG INDEX features).<br />
(43) NP-APPOS: (↓ IND) = (↑ IND)<br />
(44) X −→ X<br />
↓ ∈ ↑<br />
@NP-APPOS<br />
X<br />
↓ ∈ ↑<br />
@NP-APPOS<br />
In the <strong>in</strong>terests of clarity, we assume here that all INDEX features <strong>in</strong> appositional constructions<br />
will be shared between the members <strong>and</strong> the set. This is potentially an oversimplification, s<strong>in</strong>ce<br />
it may well be the case that there will be <strong>in</strong>stances of appositions <strong>in</strong> which the f-structures may<br />
differ <strong>in</strong> one or more INDEX features despite be<strong>in</strong>g descriptions of the same real world entity. A<br />
circumstance where this might arise could be where apparent person mismatches are allowed<br />
<strong>in</strong> appositonal structures (e.g. <strong>in</strong> the English ‘us l<strong>in</strong>guists’, ‘you children’). A further tricky<br />
area concerns gender, where a complicat<strong>in</strong>g factor <strong>in</strong> the <strong>in</strong>terpretation of appositional data is<br />
the fact that it is proposed <strong>in</strong> the literature that nouns have both INDEX GEND features (usually<br />
relevant to phenomena such as predicate argument agreement) <strong>and</strong> CONCORD GEND features<br />
(potentially relevant to agreement with<strong>in</strong> a NP, that is, cases of head modifier agreement), <strong>and</strong><br />
these may not match (Wechsler <strong>and</strong> Zlatić, 2003; K<strong>in</strong>g <strong>and</strong> Dalrymple, 2004). Well-known<br />
cases of ‘mismatch’ nouns <strong>in</strong>clude the Serbo-Croatian collective nouns of the second declension,<br />
such as deca ‘children’, which are analysed as FEM.SG CONCORD but NEUT.PL INDEX<br />
by Wechsler <strong>and</strong> Zlatić (2003). The potential for non-match<strong>in</strong>g between CONCORD <strong>and</strong> IN-<br />
DEX <strong>in</strong> GEND complicates the <strong>in</strong>terpretation of putative mismatches <strong>in</strong> appositional structures<br />
<strong>in</strong> the languages we are concerned with, because of course it may be the case that such examples<br />
<strong>in</strong>volve nouns differ<strong>in</strong>g <strong>in</strong> CONCORD GEND but not <strong>in</strong> INDEX GEND. Other cases of<br />
gender mismatch <strong>in</strong> appositional constructions could potentially arise from generic-specific<br />
constructions <strong>in</strong> which hyponyms <strong>and</strong> hypernyms clearly belong to different gender classes<br />
(e.g. VEgetable <strong>and</strong> NEuter), but we leave <strong>in</strong>vestigation of whether this occurs to further research.<br />
In sum, very little is known about gender agreement <strong>in</strong> the languages we are concerned<br />
with, <strong>and</strong> neither is the relevance of the dist<strong>in</strong>ction between INDEX <strong>and</strong> CONCORD features yet<br />
established for these languages. Should plausible examples of gender mismatch emerge, these<br />
constructions could be captured by modify<strong>in</strong>g the above analysis <strong>in</strong> a number of ways. One<br />
possibility would be to have only one daughter <strong>in</strong> the appositional phrase structure rule contribute<br />
INDEX features to the set (i.e. be associated with the NP-APPOS template above), with<br />
the INDEX features of the other daughter only partially shared, or not shared at all.<br />
17
Turn<strong>in</strong>g now to the semantics of appositional constructions, as a first approximation<br />
we take the semantics of appositional juxtapositions to be basically <strong>in</strong>tersective (apply<strong>in</strong>g to<br />
property-denot<strong>in</strong>g nom<strong>in</strong>al (rather than NP) mean<strong>in</strong>gs). One possibility is someth<strong>in</strong>g comparable<br />
to boolean <strong>and</strong> (as <strong>in</strong> the jo<strong>in</strong>t read<strong>in</strong>g of five l<strong>in</strong>guists <strong>and</strong> philosophers), tak<strong>in</strong>g two sets<br />
of properties <strong>and</strong> <strong>in</strong>tersect<strong>in</strong>g them (see Dalrymple (2004)):<br />
(45) b-<strong>and</strong> λX.λY.X ⊓ Y<br />
An alternative, which is the one we will follow here, is to model the semantics of apposition<br />
on the semantics of nom<strong>in</strong>al modification, as follows:<br />
(46) appos λQ.λP.λX.Q(X) ∧ P(X):<br />
[ ((↑ ∈)σ VAR) ⊸ ((↑ ∈) σ RESTR ) ] ⊸<br />
[ [(( ↑ ∈)σ VAR) ⊸ (( ↑ ∈) σ RESTR) ]<br />
⊸ [ ( ↑ σ VAR) ⊸ (↑ σ RESTR) ] ]<br />
On the mean<strong>in</strong>g side, this is a function which applies to two nom<strong>in</strong>al (< e, t >) mean<strong>in</strong>gs <strong>and</strong><br />
produces an abstraction over a logical conjunction of predications hold<strong>in</strong>g of this <strong>in</strong>dividual<br />
(so it takes two nom<strong>in</strong>al mean<strong>in</strong>gs <strong>and</strong> produces a nom<strong>in</strong>al mean<strong>in</strong>g, where nom<strong>in</strong>al mean<strong>in</strong>gs<br />
are of type < e, t >). On the glue side the mean<strong>in</strong>g constructor consumes one nom<strong>in</strong>al<br />
contribution <strong>and</strong> then the other nom<strong>in</strong>al contribution to produce the mean<strong>in</strong>g of the NP as a<br />
whole.<br />
We can therefore complete our analysis of appositional juxtapositions by arbitrarily associat<strong>in</strong>g<br />
the appos semantics with some daughter <strong>in</strong> the appositional phrase structure rule:<br />
(47) X −→ X<br />
↓ ∈ ↑<br />
@NP-APPOS<br />
appos<br />
X<br />
↓ ∈ ↑<br />
@NP-APPOS<br />
In order to see how this works, consider the nom<strong>in</strong>al apposition <strong>in</strong> the now familiar Wambaya<br />
example (3). The semantics associated with each of the nom<strong>in</strong>als <strong>in</strong> this construction is given<br />
<strong>in</strong> (48) <strong>and</strong> (49).<br />
(48) garidi-ni (husb<strong>and</strong>.I-ERG):<br />
λX.husb<strong>and</strong>(X): (↑ σ VAR) ⊸ (↑ σ RESTR)<br />
(49) bungmanyi-ni (old.man.I-ERG):<br />
λX.old.man(X): (↑ σ VAR) ⊸ (↑ σ RESTR)<br />
The mean<strong>in</strong>g constructor (46), associated with the “appositional” use of the juxtaposition<br />
schema, consumes (48) <strong>and</strong> (49) to produce another nom<strong>in</strong>al mean<strong>in</strong>g, as follows:<br />
18
(50) garidi-ni bungmanyi-ni ((husb<strong>and</strong>.I-ERG old.man.I-ERG):<br />
λX.old.man(X) ∧ husb<strong>and</strong>(X):<br />
(↑ σ VAR) ⊸ (↑ σ RESTR )<br />
Note that <strong>in</strong> these languages, a bare nom<strong>in</strong>al such as (48) or (49) (or <strong>in</strong>deed (50)) may be <strong>in</strong>terpereted<br />
predicatively, but may also be given a range of NP mean<strong>in</strong>gs <strong>in</strong> context (e.g. ‘the old<br />
man’, ‘an old man’, ‘old men’). Pronouns <strong>and</strong> demonstratives may occur <strong>in</strong> “determ<strong>in</strong>iz<strong>in</strong>g”<br />
function but are by no means obligatory <strong>in</strong> the production of full (referential) NP mean<strong>in</strong>gs. In<br />
these cases, where there are no demonstratives or pronouns, we take it that additional mean<strong>in</strong>g<br />
constructors (not associated with lexical material) must be available to lift nom<strong>in</strong>als <strong>in</strong>to the<br />
appropriate range of NP mean<strong>in</strong>gs. 14<br />
Other appositional constructions discussed <strong>in</strong> §2, such as generic-specific <strong>and</strong> part-whole<br />
constructions are also straghtforwardly accounted for by this analysis. These constructions are<br />
likewise licensed by the appositional phrase structure rule (47), which is fully consistent with<br />
the consensus view on the <strong>Australian</strong>ist tradition that treats such constructions as consist<strong>in</strong>g of<br />
apposed nom<strong>in</strong>als (e.g. Blake 1987, Evans 1995, Heath 1978, etc.).<br />
The f-structure correspond<strong>in</strong>g to the juxtaposed (generic-specific) construction <strong>in</strong> (52)<br />
is given <strong>in</strong> (53). St<strong>and</strong>ard nom<strong>in</strong>al lexical entries along the l<strong>in</strong>es of (48) for wanku-ya<br />
(elasmobranch-MLOC) <strong>and</strong> kulkiji-y (shark-MLOC) comb<strong>in</strong>e with the appositional mean<strong>in</strong>g<br />
constructor to give (54):<br />
(52) Dath<strong>in</strong>-a<br />
dangka-a<br />
that-NOM man-NOM<br />
kulkiji-y.<br />
shark-MLOC<br />
niya<br />
3SG.NOM<br />
wumburung-kuru raa-ja<br />
spear-PROP<br />
wanku-ya<br />
spear-ACT elasmobranch-MLOC<br />
‘That man speared a shark with a spear.’ (Evans, 1995, 244: Kayardild)<br />
14 Our account of the semantics of apposition per se must also be extended to deal with examples <strong>in</strong> which it is<br />
clear that full NPs (e.g. of type e) occur <strong>in</strong> apposition, as <strong>in</strong> the follow<strong>in</strong>g:<br />
(51) ngada bala-thu niwan-ju naljirndirri-wu, marrwa-wu niya rabi-ju<br />
I.NOM hit-POT him-MPROP scrub turkey-MPROP near-MPROP he-NOM get up-POT<br />
I’ll shoot him, the scrub turkey, he’ll fly up nearby (Evans, 1995, 239: Kayardild)<br />
We take this to be unproblematic, but leave full explication for further research.<br />
19
(53) ⎡<br />
⎢ ⎧ ⎡<br />
⎢ ⎢<br />
⎢ ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
⎣<br />
⎣<br />
⎪⎩<br />
INDEX<br />
<br />
PERS 3<br />
<br />
NUM SG<br />
PRED ‘ELASMOBRANCH’<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
PRED ‘SHARK’<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
<br />
⎤<br />
⎥<br />
⎦<br />
⎤⎫<br />
⎥<br />
⎦<br />
⎪⎬<br />
(54) wanku-ya kulkiji-y (elasmobranch-MLOC shark-MLOC)<br />
λX.elasmobranch − fish(X) ∧ shark(X): (↑ σ VAR) ⊸ (↑ σ RESTR)<br />
⎪⎭<br />
To summarise, we can account for the use of syntactic juxtaposition to encode both coord<strong>in</strong>ate<br />
<strong>and</strong> appositional constructions by mak<strong>in</strong>g two alternative sets of annotations available<br />
for the general “coord<strong>in</strong>ate” nom<strong>in</strong>al rule (29) as follows:<br />
• Annotate each dtr @NP-CNJT <strong>and</strong> some dtr @BINARY <strong>and</strong> g-<strong>and</strong> ; OR<br />
• Annotate each dtr @NP-APPOS <strong>and</strong> some dtr appos<br />
7 Inclusory Constructions<br />
In §2 above, we <strong>in</strong>troduced the <strong>in</strong>clusory construction – another type of appositional construction<br />
common to many <strong>Australian</strong> languages. In this section we show how our account of<br />
appositional constructions given above provides a natural account of <strong>in</strong>clusory constructions<br />
also. Inclusory constructions are constructions typically <strong>in</strong>volv<strong>in</strong>g two juxtaposed elements,<br />
one a pronom<strong>in</strong>al referr<strong>in</strong>g to the group as a whole <strong>and</strong> the other a (pro)nom<strong>in</strong>al pick<strong>in</strong>g out a<br />
subset of the group. 15 Examples <strong>in</strong>clude the follow<strong>in</strong>g:<br />
15 There are actually two types of <strong>in</strong>clusory construction. The first, exemplified here, <strong>in</strong>volves two juxtaposed<br />
NP elements. The second <strong>in</strong>volves a verb-coded superset pronoun <strong>and</strong> just one NP element, as <strong>in</strong> the example<br />
below. S<strong>in</strong>ger (2001) refers to these as Type 1 <strong>and</strong> Type 2 respectively.<br />
(i) nyari-bu-ydhi-ni<br />
rni-yul-pula<br />
1DU.EX-hit-RECIP-P.CON M.SG-Aborig<strong>in</strong>al-DUAL<br />
‘I <strong>and</strong> a [aborig<strong>in</strong>al] man were fight<strong>in</strong>g.’ (Heath, 1978, 291:Ng<strong>and</strong>i)<br />
Our focus here is on Type 1 <strong>in</strong>clusory constructions, s<strong>in</strong>ce they <strong>in</strong>volve the apposition of two nom<strong>in</strong>al elements.<br />
However, our analysis of Type 1 <strong>in</strong>clusory constructions can be extended to this second type through the<br />
association of the properties of the superset pronom<strong>in</strong>al with the verbal agreement morphology, <strong>in</strong>clud<strong>in</strong>g the<br />
specification that the features of the verb-<strong>in</strong>flected pronom<strong>in</strong>al be equal to the features of the set (see below). For<br />
issues of space, however, we leave exemplification of this extension of the analysis for future work.<br />
20<br />
⎤<br />
⎥<br />
⎦
(55) Tjirlpi-lu<br />
old.manERG.NAME<br />
nyupali kati-ku-nti<br />
2DU.(ERG) take-FUT-MAYBE<br />
‘You <strong>and</strong> the Old Bloke might take (us).’ (lit. ‘you two, <strong>in</strong>clud<strong>in</strong>g the old bloke, might<br />
take (us)’) (Goddard, 1985, 101: Yankunytjatjara)<br />
(56) nga-rr-a<br />
1-DU-NOM<br />
kajakaja<br />
daddyNOM<br />
warra-ja thaa-th.<br />
go-ACT<br />
return-ACT<br />
‘Daddy <strong>and</strong> I will go’ (lit. ‘we two, <strong>in</strong>clud<strong>in</strong>g daddy, will go’) (Evans, 1995, 249:<br />
Kayardild)<br />
Inclusory constructions (also called ‘plural pronoun constructions’ <strong>in</strong> the literature) are found<br />
<strong>in</strong> languages from many different families (Schwartz, 1988a,b; McNally, 1993; Lichtenberk,<br />
2000; Bril, 2004), <strong>and</strong> are described for <strong>Australian</strong> languages <strong>in</strong> S<strong>in</strong>ger (2001, 2005). The correct<br />
analysis of the <strong>in</strong>clusory construction has been the source of some debate <strong>in</strong> the literature.<br />
Schwartz (1988b,a) analyses the <strong>in</strong>clusory construction as deriv<strong>in</strong>g from a coord<strong>in</strong>ation construction,<br />
as do Hale (1966, 1973) <strong>and</strong> Nash (1986) <strong>in</strong> their analysis of <strong>in</strong>clusory constructions<br />
<strong>in</strong> the <strong>Australian</strong> language Warlpiri. S<strong>in</strong>ger (2001), on the other h<strong>and</strong>, argues that the <strong>in</strong>clusory<br />
construction <strong>in</strong> <strong>Australian</strong> languages is a dist<strong>in</strong>ct construction type, albeit similar <strong>in</strong> some<br />
respects to coord<strong>in</strong>ation constructions, based on the fact that the <strong>in</strong>clusory construction is an<br />
endocentric construction with the features of one of the elements (i.e. the superset pronom<strong>in</strong>al)<br />
be<strong>in</strong>g identical to the features of the whole argument. This is shown most clearly <strong>in</strong> a language<br />
with verbal agreement, as <strong>in</strong> the follow<strong>in</strong>g from Nunggubuyu:<br />
(57) nurru=wa-ng ma:gurn nurru<br />
we(EX.PL)killed.it [name] we(EX.PL)<br />
M <strong>and</strong> us killed it (buffalo) (Heath, 1984, 542: Nunggubuyu)<br />
In terms of their surface syntax, (Type 1) <strong>in</strong>clusory constructions generally consist of two<br />
juxtaposed nom<strong>in</strong>al elements, <strong>and</strong> are therefore syntactically similar to the other appositional<br />
constructions we have discussed above. Indeed, many language descriptions treat them as a<br />
type of appositional construction similar to part-whole <strong>and</strong>/or generic-specific constructions<br />
(e.g. Dench 1995, Evans 1995, among others). We therefore propose that they should likewise<br />
be analysed as sets at f-structure, captur<strong>in</strong>g their syntactic similarity with apposition <strong>and</strong><br />
coord<strong>in</strong>ation constructions.<br />
Inclusory constructions are a particularly <strong>in</strong>terest<strong>in</strong>g case, however, as the features of<br />
the set overall are identical to the features of one member of the set (the pronom<strong>in</strong>al), <strong>in</strong><br />
which the features of the other member must be <strong>in</strong>cluded. Thus, <strong>in</strong>clusory constructions are<br />
a composite of the coord<strong>in</strong>ation <strong>and</strong> appositional schemas presented <strong>in</strong> (42) <strong>and</strong> (47) above.<br />
The constituent correspond<strong>in</strong>g to the superset pronom<strong>in</strong>al carries the appositional template<br />
(specify<strong>in</strong>g that its INDEX features are identical to the <strong>in</strong>dex features of the whole) <strong>and</strong> the<br />
21
constituent correspond<strong>in</strong>g to the subset member carries the coord<strong>in</strong>ation template (specify<strong>in</strong>g<br />
that its INDEX features must be a subset of the INDEX features of the whole). 16<br />
(58) X −→ X<br />
↓ ∈ ↑<br />
(59) ⎡<br />
⎢ ⎧ ⎡<br />
⎢ ⎢<br />
⎢ ⎣<br />
⎢ ⎪⎨<br />
⎢<br />
⎡<br />
⎢<br />
⎢<br />
⎣<br />
⎣<br />
⎪⎩<br />
INDEX<br />
@NP-APPOS<br />
<br />
PERS 1<br />
<br />
NUM DUAL<br />
PRED ‘DADDY’<br />
INDEX<br />
NUM SG<br />
PERS 3<br />
, X<br />
↓ ∈ ↑<br />
@NP-CNJT<br />
⎤<br />
⎥<br />
⎦<br />
PRED ‘PRO’<br />
<br />
NUM DUAL<br />
INDEX<br />
PERS 1<br />
⎤<br />
⎫<br />
⎪⎬<br />
⎥<br />
⎦<br />
⎪⎭<br />
Thus, our analysis of apposition <strong>in</strong> <strong>Australian</strong> languages extends directly to <strong>in</strong>clusory constructions,<br />
correctly captur<strong>in</strong>g the fact that these various juxtaposed construction types are<br />
syntactically similar <strong>in</strong> many <strong>Australian</strong> languages. 17 Moreover, this analysis captures the<br />
various characteristics of the <strong>in</strong>clusory construction discussed <strong>in</strong> the literature. The fact that<br />
<strong>in</strong>clusory constructions are similar to coord<strong>in</strong>ation constructions is captured by the fact that,<br />
like coord<strong>in</strong>ation constructions, <strong>in</strong>clusory constructions are represented as sets at f-structure.<br />
The endocentricity of the <strong>in</strong>clusory construction (S<strong>in</strong>ger 2001), namely that the features of the<br />
whole argument are identical to those of the superset pronom<strong>in</strong>al, is captured by the fact that<br />
the pronom<strong>in</strong>al carries the apposition template, which specifies that the features of the <strong>in</strong>dividual<br />
member are carried up to the set as a whole. The <strong>in</strong>clusory nature of the construction<br />
is captured by the fact that the other element must have features which form a subset of the<br />
features of the set. 18<br />
Treat<strong>in</strong>g <strong>in</strong>clusory constructions as sets at f-structure is further supported by the fact that<br />
<strong>in</strong> some <strong>Australian</strong> languages they can be marked with suffixes that explicitly <strong>in</strong>dicate set<br />
16 This phrase structure rule is written to allow either order<strong>in</strong>g of the two elements (as <strong>in</strong>dicated by the comma<br />
between the two NPs). However <strong>in</strong> some languages it may be necessary to fix the order<strong>in</strong>g of the two elements<br />
<strong>in</strong> the case of <strong>in</strong>clusory constructions.<br />
17 The semantics of the <strong>in</strong>clusory is that one member denotes a group <strong>and</strong> the other member of the set contributes<br />
a further restriction over the group by provid<strong>in</strong>g a specification about one of its members. We leave a<br />
detailed discussion of the semantics of the <strong>in</strong>clusory for future research.<br />
18 Note that, even <strong>in</strong> languages <strong>in</strong> which the two elements can appear <strong>in</strong> either order, the nature of the templates<br />
will ensure that the apposition template is associated with the superset pronom<strong>in</strong>al <strong>and</strong> the coord<strong>in</strong>ation template<br />
with the subset (pro)nom<strong>in</strong>al. This is because, were the associations to be reversed <strong>and</strong> the coord<strong>in</strong>ation template<br />
to be associated with the superset pronom<strong>in</strong>al, its features would not be a subset of the features of the set (<strong>in</strong> this<br />
case, the features of the subset (pro)nom<strong>in</strong>al).<br />
22<br />
⎤<br />
⎥<br />
⎦
membership. In Yid<strong>in</strong>y the ‘one of a group’ suffix -ba is used <strong>in</strong> both coord<strong>in</strong>ation constructions<br />
(60) <strong>and</strong> <strong>in</strong>clusory constructions (61):<br />
(60) darnggidarnggi:ba yaburruba gal<strong>in</strong>g<br />
old.woman-ba(ABS) young.girl-ba(ABS) go-PRES<br />
‘An old woman (be<strong>in</strong>g one of a group of people) <strong>and</strong> a girl (be<strong>in</strong>g another member of a<br />
group) are go<strong>in</strong>g.’ (Dixon, 1977, 177: Yid<strong>in</strong>y)<br />
(61) nganytyi<br />
1NON-SING<br />
bunya:ba gal<strong>in</strong>g<br />
woman-ba go-PRES<br />
‘A woman <strong>and</strong> I (<strong>and</strong> some others) are go<strong>in</strong>g.’ (Dixon, 1977, 178: Yid<strong>in</strong>y)<br />
Thus, our analysis of apposition <strong>and</strong> asyndetic coord<strong>in</strong>ation <strong>in</strong> <strong>Australian</strong> languages extends<br />
naturally to <strong>in</strong>clusory constructions also, captur<strong>in</strong>g the similarities amongst these constructions<br />
types <strong>and</strong> provid<strong>in</strong>g a unified analysis of the NP juxtaposition that is so prevalent <strong>in</strong> these<br />
languages.<br />
8 Boolean <strong>Coord<strong>in</strong>ation</strong><br />
Before clos<strong>in</strong>g we turn briefly to the topic of boolean coord<strong>in</strong>ation <strong>in</strong> non-<strong>Australian</strong> languages.<br />
The conjunction of s<strong>in</strong>gular nom<strong>in</strong>als generally forms a plural noun phrase <strong>in</strong> English,<br />
under (some form of) group form<strong>in</strong>g coord<strong>in</strong>ation, but this is not always the case. As the SG<br />
verb agreement <strong>in</strong> the follow<strong>in</strong>g example <strong>in</strong>dicates, ‘vice-president’ <strong>and</strong> ‘president-elect’ refer<br />
here to the same <strong>in</strong>dividual. This boolean or jo<strong>in</strong>t read<strong>in</strong>g is <strong>in</strong> contrast to (63) which <strong>in</strong>volves<br />
a split or group-form<strong>in</strong>g read<strong>in</strong>g.<br />
(62) The vice-president <strong>and</strong> president-elect is eat<strong>in</strong>g pizza (K<strong>in</strong>g <strong>and</strong> Dalrymple, 2004, 75)<br />
(63) The vice-president <strong>and</strong> president-elect are eat<strong>in</strong>g pizza<br />
As K<strong>in</strong>g <strong>and</strong> Dalrymple (2004) show (see also Heycock <strong>and</strong> Zamparelli (1999, 2000)) languages<br />
differ <strong>in</strong> terms of the distribution of these read<strong>in</strong>g: languages such as English permit<br />
both jo<strong>in</strong>t <strong>and</strong> split read<strong>in</strong>gs (with s<strong>in</strong>gular nouns) under a s<strong>in</strong>gle determ<strong>in</strong>er such as the, or<br />
this while others permit only jo<strong>in</strong>t read<strong>in</strong>gs <strong>in</strong> these circumstances (such as Italian, Portuguese<br />
<strong>and</strong> German). The German (64), for example, has just a jo<strong>in</strong>t read<strong>in</strong>g.<br />
(64) me<strong>in</strong><br />
my-M.SG<br />
bester<br />
best<br />
Freund<br />
friend-M.SG<br />
und Mann<br />
<strong>and</strong> husb<strong>and</strong>-M.SG<br />
my best friend <strong>and</strong> husb<strong>and</strong> (K<strong>in</strong>g <strong>and</strong> Dalrymple, 2004, 93: German)<br />
23
The dist<strong>in</strong>ction between split <strong>and</strong> jo<strong>in</strong>t read<strong>in</strong>gs is most evident with s<strong>in</strong>gular conjuncts, but<br />
is also relevant <strong>in</strong> terms of the <strong>in</strong>terpretation of coord<strong>in</strong>ate structures with plural conjuncts, as<br />
<strong>in</strong> five philosophers <strong>and</strong> l<strong>in</strong>guists, which has a number of <strong>in</strong>terpretations <strong>in</strong>clud<strong>in</strong>g the jo<strong>in</strong>t<br />
read<strong>in</strong>g under which each <strong>in</strong>dividual (of 5) is both a l<strong>in</strong>guist <strong>and</strong> a philosopher.<br />
K<strong>in</strong>g <strong>and</strong> Dalrymple (2004) propose that the dist<strong>in</strong>ction between split <strong>and</strong> jo<strong>in</strong>t read<strong>in</strong>gs<br />
<strong>in</strong>volves two different semantics for <strong>and</strong>, boolean-<strong>and</strong> <strong>and</strong> group-<strong>and</strong>. For K<strong>in</strong>g <strong>and</strong> Dalrymple<br />
(2004), the former is associated with the syntactic INDEX requirement stated <strong>in</strong> (65), 19<br />
while group-form<strong>in</strong>g <strong>and</strong> simply requires a plural INDEX. The INDEX NUM feature basically<br />
encodes the number of <strong>in</strong>dividuals the phrase refers to.<br />
(65) boolean <strong>and</strong> (↑ INDEX NUM) = (↑ ∈ INDEX NUM)<br />
It is sometimes suggested that accounts of nom<strong>in</strong>al coord<strong>in</strong>ation should <strong>in</strong>volve just one semantics,<br />
<strong>in</strong> terms either of group form<strong>in</strong>g or boolean coord<strong>in</strong>ation (for example, Heycock <strong>and</strong><br />
Zamparelli (2000) propose one underly<strong>in</strong>g semantics for <strong>and</strong> <strong>and</strong> associate different (semantic)<br />
procedures with various (abstract) syntactic features <strong>in</strong> English <strong>and</strong> Italian NPs). Similarly<br />
the use of the conjunctive coord<strong>in</strong>ator <strong>and</strong> to conjo<strong>in</strong> alternative descriptions of a s<strong>in</strong>gle <strong>in</strong>dividual<br />
might seem idiosyncratic. However our proposal provides further support for the<br />
existence of both group form<strong>in</strong>g <strong>and</strong> boolean <strong>and</strong> <strong>in</strong> nom<strong>in</strong>al coord<strong>in</strong>ation by plac<strong>in</strong>g the existence<br />
of boolean coord<strong>in</strong>ation (that is, the jo<strong>in</strong>t read<strong>in</strong>g of my friend <strong>and</strong> colleague) <strong>in</strong> a rather<br />
different crossl<strong>in</strong>guistic context. We have argued that appositions, <strong>in</strong>clusories <strong>and</strong> (st<strong>and</strong>ard)<br />
conjunctions may all be associated with precisely the same syntactic device (of asyndetic coord<strong>in</strong>ation<br />
or juxtaposition). On our view, then, boolean coord<strong>in</strong>ation can be considered to be<br />
essentially similar to the appositional juxtapositions we discuss, the only difference be<strong>in</strong>g the<br />
presence of an overt coord<strong>in</strong>ator <strong>in</strong> the former (<strong>and</strong> note that <strong>in</strong> English, overt conjunctions,<br />
particularly or may occur <strong>in</strong> appositions). It is thus syntactically coord<strong>in</strong>ated (hav<strong>in</strong>g a hybrid<br />
f-structure), but semantically appositional (hav<strong>in</strong>g no feature resolution <strong>and</strong> appositional (i.e.<br />
boolean) semantics).<br />
9 Conclusion<br />
We have shown how the range of juxtaposed NP constructions <strong>in</strong> <strong>Australian</strong> languages can be<br />
accounted for relatively straightforwardly with<strong>in</strong> the constra<strong>in</strong>t-based lexicalist formalism of<br />
LFG. We have developed an account <strong>in</strong> which nom<strong>in</strong>al-nom<strong>in</strong>al sequences all have essentially<br />
the same f-structure, but correspond to 3 different feature association patterns, as <strong>in</strong> (66)-(68),<br />
<strong>and</strong> map onto a range of different semantics correlated with these three different patterns. In<br />
the case of coord<strong>in</strong>ate constructions, the INDEX features of the elements <strong>in</strong> the set-valued fstructure<br />
st<strong>and</strong> <strong>in</strong> a subset relation to the INDEX of the set itself. In the case of appositional<br />
constructions, the INDEX features of the elements are equivalent, <strong>and</strong> are also equivalent to<br />
those of the set itself. Inclusory constructions <strong>in</strong>volve one element of the set hav<strong>in</strong>g the same<br />
19 Additionally there is a semantic requirement (unformulated) which requires all conjuncts to have the same<br />
number.<br />
24
INDEX features as the set itself, while the other element <strong>in</strong> the set st<strong>and</strong>s <strong>in</strong> a subset relation to<br />
the overall INDEX.<br />
(66) coord<strong>in</strong>ation – X ⊇ Y, Z<br />
⎡<br />
⎤<br />
INDEX [X]<br />
⎢ ⎧ <br />
⎢ ⎪⎨<br />
⎢<br />
⎣ ⎪⎩<br />
<br />
INDEX [Y]<br />
INDEX [Z]<br />
⎫<br />
⎪⎬<br />
⎪⎭<br />
(67) apposition – X = Y = Z<br />
⎡<br />
INDEX [X]<br />
⎢ ⎧ <br />
⎢ ⎪⎨<br />
⎢<br />
⎣ ⎪⎩<br />
<br />
INDEX [Y]<br />
INDEX [Z]<br />
⎫<br />
⎪⎬<br />
⎪⎭<br />
(68) <strong>in</strong>clusory – X = Y ⊇ Z<br />
⎡<br />
INDEX [X]<br />
⎢ ⎧ <br />
⎢ ⎪⎨<br />
⎢<br />
⎣ ⎪⎩<br />
<br />
INDEX [Y]<br />
INDEX [Z]<br />
⎫<br />
⎪⎬<br />
⎪⎭<br />
⎥<br />
⎦<br />
⎤<br />
⎥<br />
⎦<br />
⎤<br />
⎥<br />
⎦<br />
The flexible architecture of LFG thus provides a unified syntactic account of a range of juxtaposed<br />
nom<strong>in</strong>al constructions common to <strong>Australian</strong> languages, while still captur<strong>in</strong>g their<br />
semantic differences. In this paper we have shown how the use of hybrid f-structures can be<br />
extended beyond true (semantically) coord<strong>in</strong>ated constructions to generic-specific, part-whole<br />
<strong>and</strong> other types of appositional constructions also, mak<strong>in</strong>g a dist<strong>in</strong>ction between syntactic coord<strong>in</strong>ation<br />
(hybrid structures) <strong>and</strong> semantic coord<strong>in</strong>ation (correspond<strong>in</strong>g to feature resolution<br />
<strong>and</strong> coord<strong>in</strong>ate semantics) <strong>in</strong> a simple <strong>and</strong> <strong>in</strong>tuitive way. 20 One of the implications of our analysis<br />
of <strong>Australian</strong> nom<strong>in</strong>al-nom<strong>in</strong>al constructions is that appositions are structurally the same<br />
as coord<strong>in</strong>ations <strong>in</strong> the syntax. Similar suggestions have been made <strong>in</strong> the literature, outside<br />
of the <strong>Australian</strong>ist <strong>and</strong> LFG contexts, notably <strong>in</strong> a descriptive context by Quirk et al. (1985);<br />
Meyer (1992) <strong>and</strong> others. Indeed <strong>in</strong> recent theoretical work, de Vries (2006) argues that appositive<br />
relative clauses (<strong>and</strong> other appositions) <strong>in</strong> Dutch <strong>and</strong> English should be treated as<br />
what he calles ‘specify<strong>in</strong>g coord<strong>in</strong>ation’, <strong>and</strong> analysed structurally as <strong>in</strong>volv<strong>in</strong>g CoP phrases.<br />
20 cf. Culicover <strong>and</strong> Jackendoff (1997, 2005) <strong>and</strong> Yuasa <strong>and</strong> Sadock (2002) who also discuss mismatches<br />
between ‘syntactic’ <strong>and</strong> ‘semantic’ coord<strong>in</strong>ation, although, <strong>in</strong> LFG terms their data is relevant to mismatches<br />
between c-structure <strong>and</strong> f-structure, rather than between f-structure <strong>and</strong> semantic structure, which is our concern<br />
here.<br />
25
While we are <strong>in</strong> agreement with the general thrust of the argument that apposition has much<br />
<strong>in</strong> common with coord<strong>in</strong>ation, the attempt to capture the similarities (<strong>and</strong> differences) <strong>in</strong> terms<br />
of c-structure (<strong>and</strong> with a syntax-semantics <strong>in</strong>terface which maps from c-structure to semantic<br />
structure) has a number of potentially unfortunate consequences, requir<strong>in</strong>g the postulation<br />
of null Co nodes (to head appositional (coord<strong>in</strong>ate) structures which are not marked by overt<br />
coord<strong>in</strong>ators), which our proposal avoids. Nonetheless, it is <strong>in</strong>terest<strong>in</strong>g that work <strong>in</strong> very different<br />
frameworks is converg<strong>in</strong>g on the view that apposition is, at least syntactically, a type of<br />
coord<strong>in</strong>ation.<br />
References<br />
Acuña-Fariña, Juan Carlos. 1999. On apposition. English Language <strong>and</strong> L<strong>in</strong>guistics 3(1):59–<br />
81.<br />
Aust<strong>in</strong>, Peter <strong>and</strong> Joan Bresnan. 1996. Non-configurationality <strong>in</strong> <strong>Australian</strong> Aborig<strong>in</strong>al languages.<br />
Natural Language <strong>and</strong> L<strong>in</strong>guistic Theory 14(2):215–268.<br />
Blake, Barry. 1979. A Kalkatungu Grammar. Canberra: Pacific L<strong>in</strong>guistics.<br />
Blake, Barry. 1987. <strong>Australian</strong> Aborig<strong>in</strong>al Grammar. London: Croom Helm.<br />
Blake, Barry J. 1983. Structure <strong>and</strong> word order <strong>in</strong> Kalkatungu: The Anatomy of a Flat Language.<br />
<strong>Australian</strong> Journal of L<strong>in</strong>guistics 3(2):143–175.<br />
Blake, Barry J. 2001. The Noun Phrase <strong>in</strong> <strong>Australian</strong> <strong>Languages</strong>. In Jane Simpson et. al.,<br />
ed., Forty years on: Ken Hale <strong>and</strong> <strong>Australian</strong> languages, pages 415–425. Canberra: Pacific<br />
L<strong>in</strong>guistics.<br />
Bresnan, Joan. 2001. Lexical Functional Syntax. Oxford: Blackwell.<br />
Bril, Isabelle. 2004. <strong>Coord<strong>in</strong>ation</strong> Strategies <strong>and</strong> Inclusory Constructions <strong>in</strong> New Caledonian<br />
<strong>and</strong> other Oceanic <strong>Languages</strong>. In M. Haspelmath, ed., Coord<strong>in</strong>at<strong>in</strong>g Constructions, pages<br />
499–533. Amsterdam: John Benjam<strong>in</strong>s.<br />
Crystal, David. 1997. A Dictionary of L<strong>in</strong>guistics <strong>and</strong> Phonetics. Oxford: Blackwell.<br />
Culicover, Peter <strong>and</strong> Ray Jackendoff. 1997. Semantic subord<strong>in</strong>ation despite syntactic coord<strong>in</strong>ation.<br />
L<strong>in</strong>guistic Inquiry 28:195–218.<br />
Culicover, Peter <strong>and</strong> Ray Jackendoff. 2005. Simpler Syntax. Oxford: Blackwells.<br />
Dalrymple, Mary. 2001. Lexical Functional Grammar. San Diego, CA: Academic Press.<br />
Dalrymple, Mary. 2004. Noun coord<strong>in</strong>ation: Syntax <strong>and</strong> semantics. talk given at the University<br />
of Canterbury, New Zeal<strong>and</strong>.<br />
26
Dalrymple, Mary, Helge Dyvik, <strong>and</strong> Louisa Sadler. 2007. Gender resolution <strong>and</strong> gender <strong>in</strong>determ<strong>in</strong>acy.<br />
LFG Conference 2007.<br />
Dalrymple, Mary, Ron Kaplan, <strong>and</strong> Tracy Holloway K<strong>in</strong>g. 2004. L<strong>in</strong>guistic generalizations<br />
over descriptions. In M. Butt <strong>and</strong> T. H. K<strong>in</strong>g, eds., Proceed<strong>in</strong>gs of the LFG04 Conference.<br />
Stanford, CA: CSLI Publications: http://www-csli.stanford.edu/publications.<br />
Dalrymple, Mary <strong>and</strong> Ronald M. Kaplan. 2000. Feature <strong>in</strong>determ<strong>in</strong>acy <strong>and</strong> feature resolution.<br />
Language 76(4):759–798.<br />
Dalrymple, Mary, Ronald M. Kaplan, John T. Maxwell, III, <strong>and</strong> Annie Zaenen, eds. 1995.<br />
Formal Issues <strong>in</strong> Lexical-Functional Grammar. Stanford, CA: CSLI Publications.<br />
de Vries, Mark. 2006. The syntax of appositive relativization: on specify<strong>in</strong>g coord<strong>in</strong>ation, free<br />
relatives <strong>and</strong> promotion. L<strong>in</strong>guistic Inquiry 37:229–270.<br />
Dench, Alan. 1995. Martuthunira: a language of the Pilbara Region of Western Australia.<br />
Canberra: Pacific L<strong>in</strong>guistics.<br />
Dixon, R. M. W. 1977. A grammar of Yid<strong>in</strong>y. Cambridge: Cambridge University Press.<br />
Evans, Nicholas. 1995. A Grammar of Kayardild: With Historical-Comparative Notes on<br />
Tangkic. Berl<strong>in</strong>: Mouton de Gruyter.<br />
Falk, Yehuda. 2001. Lexical-Functional Grammar: An Introduction to Parallel Constra<strong>in</strong>t-<br />
Based Syntax. Stanford, CA: CSLI Publications.<br />
Gaby, Alice. 2006. A Grammar of Kuuk Thaayorre. Ph.D. thesis, University of Melbourne,<br />
Melbourne, Australia.<br />
Goddard, Clifford. 1985. A grammar of Yankunytjatjara. Alice Spr<strong>in</strong>gs: Institute for Aborig<strong>in</strong>al<br />
Development.<br />
Hale, Ken. 1966. K<strong>in</strong>ship reflections <strong>in</strong> syntax. Word 22:318–324.<br />
Hale, Ken. 1973. Person mark<strong>in</strong>g <strong>in</strong> Warlbiri. In S. Anderson <strong>and</strong> P. Kiparsky, eds., A<br />
Festschrift for Morris Halle. New York: Holt, R<strong>in</strong>ehard <strong>and</strong> W<strong>in</strong>ston.<br />
Heath, Jeffrey. 1978. Ng<strong>and</strong>i grammar, texts <strong>and</strong> dictionary. Canberra: Pacific L<strong>in</strong>guistics.<br />
Heath, Jeffrey. 1984. Functional Grammar of Nunggubuyu. Canberra: AIAS.<br />
Heycock, Carol<strong>in</strong>e <strong>and</strong> Roberto Zamparelli. 1999. Toward a unified analysis of DP conjunction.<br />
In P. Dekker, ed., Proceed<strong>in</strong>gs of the Twelfth Amsterdam Colloquium, pages 127–132.<br />
ILLC, University of Amsterdam.<br />
Heycock, Carol<strong>in</strong>e <strong>and</strong> Roberto Zamparelli. 2000. Friends <strong>and</strong> colleagues: Plurality <strong>and</strong> NPcoord<strong>in</strong>ation.<br />
In M. Hirotani, A. Coetzee, N. Hall, <strong>and</strong> J.-Y. Kim, eds., Proceed<strong>in</strong>gs of the<br />
30th Annual Meet<strong>in</strong>g of the North Eastern L<strong>in</strong>guistic Society. Amherst, MA: GLSA.<br />
27
Kaplan, Ronald M. <strong>and</strong> John T. Maxwell, III. 1988. Constituent coord<strong>in</strong>ation <strong>in</strong> Lexical-<br />
Functional Grammar. In Proceed<strong>in</strong>gs of the 12th International Conference on Computational<br />
L<strong>in</strong>guistics (COLING88), Budapest, vol. 1, pages 303–305. Repr<strong>in</strong>ted <strong>in</strong> Dalrymple<br />
et al. (1995, pp. 199–210).<br />
K<strong>in</strong>g, Tracy Holloway <strong>and</strong> Mary Dalrymple. 2004. Determ<strong>in</strong>er agreement <strong>and</strong> noun conjunction.<br />
Journal of L<strong>in</strong>guistics 40(1):69–104.<br />
Lichtenberk, Frantisek. 2000. Inclusory pronom<strong>in</strong>als. Oceanic L<strong>in</strong>guistics 39(1):1–32.<br />
McNally, Louise. 1993. Comitative <strong>Coord<strong>in</strong>ation</strong>: A Case Study <strong>in</strong> Group Formation. Natural<br />
Language <strong>and</strong> L<strong>in</strong>guistic Theory 11:347–79.<br />
Meyer, Charles F. 1992. <strong>Apposition</strong> <strong>in</strong> contemporary English. Cambridge: Cambridge University<br />
Press.<br />
Nash, David. 1986. Topics <strong>in</strong> Warlpiri Grammar. Ph.D. thesis, MIT.<br />
Nordl<strong>in</strong>ger, Rachel. 1998a. A Grammar of Wambaya, Northern Territory (Australia). Canberra:<br />
Pacific L<strong>in</strong>guistics.<br />
Nordl<strong>in</strong>ger, Rachel. 1998b. Constructive Case: Evidence from <strong>Australian</strong> <strong>Languages</strong>. Stanford,<br />
CA: CSLI Publications.<br />
Nordl<strong>in</strong>ger, Rachel. 1998c. Constructive Case: Evidence from <strong>Australian</strong> <strong>Languages</strong>. Stanford,<br />
CA: CSLI Publications.<br />
Pensalf<strong>in</strong>i, Robert. 2003. A grammar of J<strong>in</strong>gulu: an Aborig<strong>in</strong>al language of the Northern<br />
Territory. Canberra: Pacific L<strong>in</strong>guistics.<br />
Quirk, R<strong>and</strong>olph, Sidney Greenbaum, Geoffrey Leech, <strong>and</strong> Jan Svartvik. 1985. A Comprehensive<br />
Grammar of the English Language. London: Longman.<br />
Schwartz, L<strong>in</strong>da. 1988a. Asymmetric Feature Distribution <strong>in</strong> Pronom<strong>in</strong>al ‘<strong>Coord<strong>in</strong>ation</strong>s’.<br />
In M. Barlow <strong>and</strong> C. A. Ferguson, eds., Agreement <strong>in</strong> Natural Language, pages 237–50.<br />
Stanford: CSLI Publications.<br />
Schwartz, L<strong>in</strong>da. 1988b. Conditions for verb-coded coord<strong>in</strong>ation. In M. Hammond,<br />
E. Moravcsik, <strong>and</strong> J. Wirth, eds., Studies <strong>in</strong> Syntactic Typology, pages 53–73. Amsterdam:<br />
John Benjam<strong>in</strong>s.<br />
Sharp, Janet. 2004. Nyangumarta: A Language of the Pilbara Region of Western Australia.<br />
Canberra: Pacific L<strong>in</strong>guistics.<br />
Simpson, Jane. 1991. Warlpiri Morpho-Syntax. Dordrecht: Kluwer Academic Publishers.<br />
Simpson, Jane <strong>and</strong> Joan Bresnan. 1983. Control <strong>and</strong> Obviation <strong>in</strong> Warlpiri. Natural Language<br />
<strong>and</strong> L<strong>in</strong>guistic Theory 1:49–64.<br />
28
S<strong>in</strong>ger, Ruth. 2001. The <strong>in</strong>clusory construction <strong>in</strong> australian languages. Honours thesis, University<br />
of Melbourne.<br />
S<strong>in</strong>ger, Ruth. 2005. Compar<strong>in</strong>g constructions across languages: a case study of the relationship<br />
between the <strong>in</strong>clusory construction <strong>and</strong> some related nom<strong>in</strong>al constructions. Unpublished<br />
talk, ALT 2005.<br />
Van Eynde, Frank. 2005. A head-driven treatment of asymmetric coord<strong>in</strong>ation <strong>and</strong> apposition.<br />
In S. Müller, ed., Proceed<strong>in</strong>gs of HPSG 2005, pages 396–409. CSLI Publications:<br />
http://www-csli.stanford.edu/publications, Stanford, CA.<br />
Wechsler, Stephen <strong>and</strong> Larisa Zlatić. 2003. The Many Faces of Agreement. Stanford, CA:<br />
CSLI Publications.<br />
Yuasa, Etsuyo <strong>and</strong> Jerry Sadock. 2002. Pseudo-subord<strong>in</strong>ation: a mismatch between syntax<br />
<strong>and</strong> semantics. Journal of L<strong>in</strong>guistics 38:87–111.<br />
29